Combine PDF

Available online at www.sciencedirect.
com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Transportation
Available Research
online Procedia 00 (2018) 000–000
at www.sciencedirect.com
Transportation Research Procedia 00 (2018) 000–000 www.elsevier.com/locate/procedia
ScienceDirect www.elsevier.com/locate/procedia
Transportation Research Procedia 39 (2019) 300–308

www.elsevier.com/locate/procedia
Green Cities 2018

Green Cities 2018
Modeling traffic flow on two-lane roads with traffic lights and
Modeling traffic flow on two-lane timer
countdown roads with traffic lights and
countdown timer
Krzysztof Małeckia,*
a,*
, Stanisław Iwanbb
a
Krzysztof Małecki , Stanisław Iwan
West Pomeranian University of Technology, Żołnierska Str. 52, 71-210 Szczecin, Poland
b
Maritime University of Szczecin,
a Faculty of University
West Pomeranian Economicsofand EngineerinŻołnierska
Technology, of Transport, 11 Henryka
Str. 52, Pobożnego
71-210 Szczecin, Str. 70-507 Szczecin, Poland
Poland
b
Maritime University of Szczecin, Faculty of Economics and Engineerin of Transport, 11 Henryka Pobożnego Str. 70-507 Szczecin, Poland
Abstract
Abstract
The paper presents the analysis of the impact of various drivers' behaviours on the traffic flow in the city. The analysis of the
countdown timers mounted
The paper presents at traffic
the analysis of thelights andoftheir
impact positive
various influence
drivers' on traffic
behaviours flow
on the was also
traffic flow presented.
in the city.The
Thestudy was of
analysis based
the
on a simulation
countdown system
timers mountedbasedat on cellular
traffic automaton
lights and their(CA) model
positive and multi-agent
influence on traffic technology.
flow was also Thepresented.
developed CAstudy
The modelwasis based
on acells of smaller
simulation size based
system than theon size of the
cellular CA cell (CA)
automaton in themodel
Nagel-Schreckenberg traffic model.
and multi-agent technology. TheThe model also
developed CAreflects
model is thebased
real
behaviour
on cells of of driverssize
smaller described
than theinsize
theofliterature
the CA and
cell observed in the reality. The traffic
in the Nagel-Schreckenberg research experiments
model. The modelwere carried
also outthe
reflects on real
the
example
behaviourofof a two-lane road withintraffic
drivers described lights. and observed in the reality. The research experiments were carried out on the
the literature
example of a two-lane road with traffic lights.
© 2018 The Authors. Published by Elsevier B.V.
© 2019 The Authors. Published by Elsevier B.V.
This
© is an
2018 open
The accessPublished
Authors. article under the CC BY-NC-ND
by Elsevier B.V. license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
Selection
This is an and
open peer-review
access under
article underresponsibility
the CC of the
BY-NC-ND
Selection and peer-review under responsibility of the scientificscientific
licensecommittee of
of Green
Green Logistics
Logistics for
for Greener
Greener Cities
Cities 2018.
(https://creativecommons.org/licenses/by-nc-nd/4.0/)
committee 2018.
Selection and peer-review under responsibility of the scientific committee of Green Logistics for Greener Cities 2018.
Keywords: Cellular Automata (CA); computer simulation; traffic modelling; traffic lights countdown timer
Keywords: Cellular Automata (CA); computer simulation; traffic modelling; traffic lights countdown timer
1. Introduction
1. Introduction
Almost all road intersections are organized on the basis of traffic lights that control the flow of vehicles and
Almost allThis
pedestrians. roadis intersections are organized
a good and practical on the
solution basis
since roadoftraffic
trafficvaries
lights greatly
that control the flow
over time. of vehicles
In many cases, and
the
pedestrians.
traffic lights This
reduce is the
a good and practical
flexibility of trafficsolution
flow. Tosince
makeroad traffic
traffic roadvaries
more greatly over time.
fluent, various In many
scenarios andcases,
methodsthe
for switching
traffic lights at
lights reduce theintersection aretraffic
flexibility of developed
flow. (Małecki, Pietruszka
To make traffic road & Iwan,
more 2017;various
fluent, Małecki & Pietruszka,
scenarios 2018)
and methods
for
and switching lightsofat traffic
adaptive type intersection
lights are developed (Małecki,
is introduced (Touhbi etPietruszka
al., 2017).& AnIwan, 2017; Małecki
additional element&that Pietruszka,
improves2018)
this
and adaptive
situation is thetype of traffic
drivers' abilitylights is introduced
to fluent start when(Touhbi et al.,
the green light2017).
turns An
on. additional
The ideal element
situation that improves
is when, after this
the
situation is the drivers' ability to fluent start when the green light turns on. The ideal situation is when, after the
*
Corresponding author. Tel./fax: +48-91-449-5661
*
E-mail address:
Corresponding kmalecki@wi.zut.edu.pl
author. Tel./fax: +48-91-449-5661
E-mail address: kmalecki@wi.zut.edu.pl
2352-1465 © 2018 The Authors. Published by Elsevier B.V.
This is an open
2352-1465 access
© 2018 Thearticle under
Authors. the CC BY-NC-ND
Published license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
by Elsevier B.V.
Selection
This is an and
openpeer-review under
access article responsibility
under of the scientific
the CC BY-NC-ND licensecommittee of Green Logistics for Greener Cities 2018.
(https://creativecommons.org/licenses/by-nc-nd/4.0/)
2352-1465  2019 The Authors. Published by Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
10.1016/j.trpro.2019.06.032
Krzysztof Małecki et al. / Transportation Research Procedia 39 (2019) 300–308 301
2 Krzysztof Małecki Stanisław Iwan / Transportation Research Procedia 00 (2018) 000–000
green light turns on, all drivers start moving fluently, almost simultaneously, one after the other. Unfortunately, the
observation of road traffic shows that it is repeatedly different and, for various reasons, some drivers move with a
certain delay, limiting the capacity of a given phase of green light at the intersection and increasing the level of local
environmental pollution (Małecki, 2016).
The purpose of this article is to develop a multi-agent system to study the capacity of a road in relation to the
various behaviours of drivers starting moving when the green light is switched on. A traffic simulator was
developed, on the basis of which numerous experimental studies were carried out. A two-layer model has been
developed: one layer is an agent-based model (agents represent drivers with certain behaviours) and a grid of
Cellular Automaton (CA) (road representation, allowing vehicles to move, in accordance with existing traffic
dynamics models)Here introduce the paper, and put a nomenclature if necessary, in a box with the same font size as
the rest of the paper. The paragraphs continue from here and are only separated by headings, subheadings, images
and formulae. The section headings are arranged by numbers, bold and 10 pt. Here follows further instructions for
authors.
2. Related work
Road traffic analysis and modelling is a challenging task due to the complexity of the situation and its stochastic
nature. Attempting to avoid inconveniencing the connected driver with attempts to adjust a specific area (e.g., an
intersection or a roundabout), road administrators and researchers resort to simulation to determine best road
solutions. The literature describes mathematical models, software solutions, hardware solutions, as well as solutions
combining hardware and software. The most numerous are software solutions, for example: Multi-Agent Transport
Simulation (Balmer et al., 2009), Verkehr In Städten-SIMulationsmodell (Fellendorf, 1994), TRansportation
ANalysis SIMulation System (Smith et al., 1995), Massachusetts Institute of Technology SIMulation (Yang &
Koutsopoulos, 1996), Advanced Interactive Micro-scopic Simulator for Urban Networks (Barceló & Ferrer, 1997),
Simulation of Urban MObility (Krajzewicz et al., 2002), simulator dedicated to roundabouts (Małecki, 2018) and
multi-agent simulations (Chmielewska, Kotlarz & Wąs, 2016). The available literature describes also hardware
solutions (e.g. Han & Sun, 2003; Bullock & Urbanik, 2000; Jaszczak & Małecki, 2013), models (e.g. Sierpiński,
2012; Macioszek, 2016, Macioszek, 2018), traffic analyses (Karoń & Mikulski, 2011; Nowosielski et al., 2016;
Macioszek, 2018a). Sobota et al. (Sobota, Klos & Karoń, 2017) indicate the use of countdown timers in the context
of vehicle and pedestrian traffic safety.
The simple and effective computational method of traffic was proposed by Nagel and Schreckenberg (Nagel &
Schreckenberg, 1992). The authors formulated a stochastic model of cellular automaton for single-lane roads
(NaSch model). In the developed model time and space are discretized, and due to low computational complexity,
the model can be used to model a large number of vehicles (Hoogendoorn & Bovy, 2001). The NaSch model was
extended for two-lane traffic model (Nagel et al., 1998). Traffic flow was considered in many articles (e.g. Biham,
Middleton & Levine, 1992; Chopard, Luthi & Queloz, 1996; Schadschneider & Schreckenberg, 1997; Chowdhury
& Schadschneider, 1999; Chowdhury, Santen & Schadschneider, 2000; Małecki & Iwan, 2012; Gwizdałła &
Grzebielucha, 2010, Małecki, 2017). The impact of the location of loading/unloading bays on the road and their
impact on the environment have been also studied (Iwan & Małecki, 2017; Iwan et al., 2018). The ability to manage
with various types of traffic and high computational efficiency made CA a widely used tool for simulating traffic’s
stochastic nature (Małecki & Wątróbski, 2017; Małecki, 2017a).
Zamith et al. (Zamith, Leal-Toledo, Clua & Toledo, 2015) presented a model that takes into account the driver's
reaction to the vehicle in front of the car. The authors used the probability density function to model behaviours of
various drivers. The test was based on three different driver behaviours: a courageous driver, slow (as opposed to a
courageous driver) and a standard driver.
The models and drivers' behaviour discussed in the literature and own observations were the basis for further
work presented in this article.
302 Krzysztof Małecki et al. / Transportation Research Procedia 39 (2019) 300–308
Krzysztof Małecki Stanisław Iwan / Transportation Research Procedia 00 (2018) 000–000 3
3. Proposed approach
The CA paradigm stems from the attempts made by scientists to develop seemingly complicated processes in the
form of a series of simple local decisions (Wolfram, 2002). Implementation of the paradigm requires the
presentation of analysed space in the form of a homogeneous network of cells. The decision at the cell level is made
on the basis of the transition function, which depends on the analysed cell and its surroundings, i.e. cells adjacent to
it. Each cell takes one state from a finite set of states. The network can have any number of finite sizes. However,
one- or two-dimensional are the most commonly used. The time in the model is discreet. Adjacent cells are
connected to a specific cell and they do not change (Mamei, Roli, & Zambonelli, 2005). Due to the iterative use of
rules, the CA process coincides with the description of global system behaviour (Missoum, Gürdal, & Setoodeh,
2005). The CA is considered as a mature computational system (Wąs & Sirakoulis, 2015).
The proposed model is a discrete, non-deterministic rule-based asynchronous model. It is a two-layer model, in
which the first layer (the layer of the road) is a CA grid describing the road. The second layer describes agents
(vehicles) who perform certain specific tasks. The CA grid describing a one-lane, one-way road is known from the
Nagel-Schreckenberg (NaSch) model (Nagel & Schreckenberg, 1992), and the two-lane road is known from [e.g.,
Nagel, Wolf, Wagner, & Simon, 1998; Rickert, Nagel, & Schreckenberg, 1996; Chowdhury & Wolf, 1997; Wang &
Zhou, 2008; Zhu, Zhang, & Wu, 2015). All these works are based on the NaSch model, which was originally
designed for freeway traffic. By reducing the size of the CA cell to 2.5m, here presented model has been adapted to
the urban conditions (Table 1), according to speed formula. The developed system updates every one second.
Table 1. Comparison of velocity in the NaSch model and the developed model presented in this article (the
difference in values results from a different cell size in both models). The values are computed based on speed
formula.
Speed [cells/iteration of CA] Speed in NaSch model [km/h] Speed in presented model [km/h]
0 0 0
1 26.8 8.9
2 53.6 17.9
3 80.4 26.8
4 107.1 35.7
5 133.9 44.6
6 160.7 53.6
In a developed CA model, each cell may be in one of the states: empty or occupied. The cell size corresponds to
the actual average length of the passenger vehicle and the physical width of the lane. The considered road section
consists of two lanes with a width of 3 meters each. It was assumed that vehicles only go in one direction. The
analysed road, on which the traffic light is placed, is 300m long, and the traffic light is located on 250m from the
beginning of the road (Fig. 1). The developed simulation system makes it possible to regulate the length of the road
depending on the simulation needs. Additionally, the red and green light phase lengths can be changed, the traffic
light device can be placed in different location, and an additional display (countdown timer) indicating the time left
to switch the lights (from red to green and from green to red) can be (or not) localised on the road.
In the proposed model, the vehicle is an agent that occupies several cells. The inspiration for this approach came
from another study (Hartman, 2004; Chmielewska, Kotlarz, & Wąs, 2016). Each car is divided into smaller parts: a
head (front of the vehicle) and a tail (other fragments). The head determines the position of the vehicle on the road;
the tail determines the length of the vehicle, which allows for mapping various types of vehicles. For a passenger
vehicle with a length of approximately 4.5m, two cells are occupied; for a freight vehicle with a length of approx.
7.8m, three CA cells; for a heavy goods vehicle (HGV) of 9.8 m length, four cells are occupied, etc.
Fig. 1. Schematic sketch of the road segment without countdown timer (top) and with countdown timer (bottom). White cells denote the one-way
double-lane street, while blue rectangles represent cars.
Partially following (Zamith, Leal-Toledo, Clua, & Toledo, 2015), the developed model takes into account the
different drivers' behaviour (Fig. 2):
 the carefree driver - a driver who is thoughtful and does not watch the traffic light changes to green and therefore
starts with a delay (e.g. he/she starts when he/she hears a horn or when he/she sees that the vehicle is moving on
an adjacent lane, etc.). Possible values of the representing variable are yes or no (represent by one or zero);
 the driver with little skills - a driver who cannot start moving fluently with the vehicle in front of. In that
situation, it happens that when car starts the engine will turn off or the vehicle will jerk and brake, and the vehicle
does not increase the speed in the expected time. Possible values of the representing variable are yes or no
(represent by one or zero);
 the standard driver - a driver who is careful and he/she has a lot of driving skills;
In addition, the developed simulator allows setting a constant delay in moving when green light turns on. The
delay is represent by variable with the set of value: from one to five.
Fig. 2. Green rectangles represent cars with drivers with little skills, red – cars with carefree drivers, while blue – standard drivers.
4. The system developed for the simulation
A computer program was developed to simulate various phenomena connected with drivers' behaviour in road
traffic. The application consists of two main components. The first one is the computational layer in which traffic
models have been implemented. This component was developed in C++. The second component is the visualisation
layer, which constitutes an interface between the application and the user. This part of the application was developed
in QML, being the main part of Qt Quick technology. The communication between the layers was provided using
the ready-made interfaces and mechanisms implemented in Qt.
AreaModel CarsModel
ObstacleAreaModel ParkingLaneAreaModel
LightsAreaModel TrucksModel
Fig. 3. Class diagram.
The application was developed using the object-oriented paradigm. Fig. 3 shows the structure of the
computational layer of the application. The inheritance mechanism and polymorphism were applied, which in
consequence led to a reduction in the code amount. Each of the classes has its own intended use:
 AreaModel – this is the basic class which includes a table representing the cellular automata grid and
implementing the NaSch traffic model.
 ObstacleAreaModel – is a derivative class which extends the model to include behaviours connected with an
obstacle on the road,
 ParkingLaneAreaModel – is a derivative class which extends the model to include behaviours connected with on-
street parking,
 LighsAreaModel – is a derivative class which extends the model to include behaviours connected with traffic
lights,
 TrucksModel – is a derivative class which extends the model to include the possibility of freight traffic and driver
behaviours connected with it,
 CarsModel – is a model developed in accordance with Qt characteristics to store all the vehicles from the cellular
automaton currently used for simulation, it enables visualisation of vehicles and their movement, using ready-
made Qt Quick mechanisms.
The application is a continuously developed tool. In its present form, it consists of four modules that make it
possible to carry out research in the area of various simulation scenarios connected with examining road traffic
effectiveness in the following aspects: an obstacle found in a lane, on-street parking, traffic lights countdown timers
and impact of trucks on the traffic. Fig. 4 shows the main window of the developed application.
Fig. 4. The main window of the developed road traffic simulator.

Fig. 5. The view window for Scenario 3 simulation – the traffic lights countdown timer module.
Fig. 5 shows the main window of the developed application and windows for the traffic lights countdown timer
module.
The developed system enables the user to input a few initial values, being the parameters for the model: road
length, number of lanes on the road, location of traffic lights, length of the green and red light phase. In addition, on
the visualization screen, the application allows changing the parameters affecting the simulation: the probability of
new vehicle appearance, the probability of generating a carefree driver, the probability of generating a driver with
low skills and the delay time in moving when green light is turn on.
5. Experimental results
The main purpose of this article was to develop a cellular automata model, aimed at investigating the impact of
delays in starting from traffic lights on road traffic capacity.
Many research experiments have been carried out taking into account different drivers' behaviour, of which a few
are presented in this chapter. Figures 5-7 show the average number of vehicles passing through green light, taking
into account the changing number of drivers with low skills, carefree drivers and the existence (or not) of
countdown timers mounted at traffic lights. The duration of the green light phase was set to 20 seconds and red to 10
seconds, in described experiments.
Fig. 6. The impact of drivers with little skill on traffic flow.
Fig. 6 shows the impact of the drivers with little skill on the traffic flow, assuming that the lack of driver's
abilities when moving off causes some delays of 1, 3 and 5 seconds. The value analysis shows that the number of
vehicles decreases with the size of the delay. If every moving vehicle is delayed by 5 seconds, then in the assumed
phase of green light the number of vehicles passing the green light is less by 20 vehicles (in relation to 26 vehicles
passing during the green light phase, with no delay and assuming that all drivers have the appropriate skills). The
experiment showed how important are the driver's abilities in context of starting car.
Fig. 7. The impact of carefree drivers on traffic flow.
Another experiment concerned the study of the effects of carefree drivers on traffic flow (Fig. 7). It turns out that
a carefree driver has little more negative impact on traffic flow than a driver with little skills. Both groups of drivers
cause significant delays in the flow of vehicles and thus increase the phenomenon of traffic congestion.
Fig. 8. The impact of traffic lights countdown timer on traffic flow.
The last experiment presented in the article shows the positive effect of countdown timers mounted at traffic
lights on vehicle flow (Fig. 8). This can be explained by the fact that drivers focus more on the countdown timer
than on the surrounding, which means that they are more ready to move at the right time. As a result, a larger flow
of vehicles is obtained on the road equipped with traffic lights (at crossings or intersections).
It is worth mentioning here that the greater the delays in starting individual vehicles from lights are, the greater
the phenomenon of congestion is and the greater local pollution of the environment is.
6. Conclusion
This paper is focused on applying computer simulation in modelling a traffic flow on two-lane roads with traffic
lights in relation to different drivers’ behaviour. A model was developed and implemented in the form of a
simulation system, which served to carry out the research study described herein.
The basic CA traffic model has been modified in the context of the size of the CA cell, which contributed to
achieving lower vehicle speeds and allowed for a more realistic mapping of vehicle traffic in the city. In addition,
various dependencies have been introduced to symbolize the different behaviour of drivers who move when green
light comes on. Countdown timers have been introduced to count down the time until green light turns on.
Experimental studies were carried out and the results are shown in the figures. Positive influence of countdown
timer has been demonstrated and it has been shown that deficiencies in driving skills and carelessness in guarding
the change of lights translate into delays in road traffic and thus for greater congestion and greater local
environmental pollution.
This paper should help educate drivers by identifying negative factors in drivers’ behaviour. It is also a basis for
installing countdown timers whose positive influence is shown in the research experiment.
Further work will focus on a more accurate representation of the processes taking place when drivers starting
from traffic lights mounted at intersections of roads or at pedestrian crossings.
References
Balmer, M., Rieser, M., Meister, K., Charypar, D., Lefebvre, N., & Nagel, K. (2009). MATSim-T: Architecture and simulation times. In: Multi-
Agent Systems for Traffic and Transportation Engineering; IGI Global: Hershey, PA, USA, p. 57–78.
Barceló, J.; & Ferrer, J. L. (1997). AIMSUN2: Advanced Interactive Microscopic Simulator for Urban Networks, User’s Manual. Universidad
Politécnica de Cataluña: Barcelona, Spain.
Biham, O., Middleton, A. A., & Levine D. (1992). Self-organization and a dynamical transition in traffic-flow models. Phys. Rev. A, 46(10),
6124.
Bullock, D., & Urbanik, T. (2000). Hardware-in-the-loop evaluation of traffic signal systems. In Proceedings of the 10th International Conference
on Road Transport Information and Control, London, UK, 4–6 April 2000, p. 177–181.
Chmielewska M., Kotlarz M., & Wąs J. (2016). Computer Simulation of Traffic Flow Based on Cellular Automata and Multi-agent System. In:
Wyrzykowski R., Deelman E., Dongarra J., Karczewski K., Kitowski J., Wiatr K. (eds) Parallel Processing and Applied Mathematics. LNCS,
vol. 9574, p. 517-527. Springer, Cham.
Chopard, B., Luthi, P. O., & Queloz, P. (1996). Cellular automata model of car traffic in a two-dimensional street network. J. Phys. A, 29(10),
2325–2336.
Chowdhury, D., & Schadschneider, A. (1999). Self-organization of traffic jams in cities: effects of stochastic dynamics and signal periods.
Physical Review E, 59, 1311–1314.
Chowdhury, D., Santen, L., & Schadschneider, A. (2000). Statistical physics of vehicular traffic and some related systems, Physics Reports,
329(4–6), 199–329.
Chowdhury, D., Wolf, D. E. (1997). Praticle hopping models for two-lane traffic with two kinds of vehicles: Effects of lane-changing rules.
Physica A Stat. Mech. Appl., 235, 417–439.
Fellendorf, M. (1994). VISSIM: A microscopic simulation tool to evaluate actuated signal control including bus priority. In Proceedings of the
64th Institute of Transportation Engineers Annual Meeting, Dallas, TX, USA, 16–19 October 1994, Technical Paper. p. 1–9.
Gwizdałła, T. M., Grzebielucha, S. (2010). The traffic flow through different form of intersections, In: Computer Information Systems and
Industrial Management Applications (CISIM), Int. Conf., IEEE, 299–304.
Han, X., & Sun, H. (2003). The Implementation of Traffic signal Light Controlled by PLC. J. Changchun Inst. Opt. Fine Mech., 4, 29.
Hartman, D. (2004). Head leading algorithm for urban traffic modeling. In: 16th International European Simulation Symposium and Exhibition
ESS, p. 297–302.
Hoogendoorn, S., & Bovy, P. (2001). State-of-the-art of vehicular traffic flow modelling. J. Syst. Control Eng., 215(4), 283–303.
Iwan, S., & Małecki, K. (2017). Utilization of cellular automata for analysis of the efficiency of urban freight transport measures based on
loading/unloading bays example. Transportation Research Procedia, 25, 1021–1035.
Iwan, S., Kijewska, K., Johansen, B. G., Eidhammer, O., Małecki, K., Konicki, W., & Thompson, R.G. (2018). Analysis of the environmental
impacts of unloading bays based on cellular automata simulation. Transportation Research Part D: Transport and Environment, 61(A), 104–
117.
Jaszczak, S., & Małecki, K. (2013). Hardware and software synthesis of exemplary crossroads in a modular programmable controller. Przegląd
Elektrotechniczny, 89(11), 121–124.
Karoń, G., & Mikulski, J. (2011). Transportation Systems Modelling as Planning, Organisation and Management for Solutions Created with ITS.
In: J. Mikulski (ed.) Modern Transport Telematics, CiCIS 239, 277–290. Springer-Verlag Berlin Heidelberg.
Krajzewicz, D., Hertkorn, G., Rössel, C., & Wagner, P. (2002). SUMO (Simulation of Urban MObility)—An open-source traffic simulation. In
Proceedings of the 4th Middle East Symposium on Simulation and Modelling (MESM2002), Berlin-Adlershof, Germany, 1–30 September
2002, p. 183–187.
Macioszek E. (2018). The comparison of models for follow-up headway at roundabouts. In: E. Macioszek, G. Sierpiński (eds.) Recent Advances
in Traffic Engineering for Transport Networks and Systems. LNNS 21. Springer International Publishing Switzerland, p. 16–26.
Macioszek, E. (2016). The application of HCM 2010 in the determination of capacity of traffic lanes at turbo roundabout entries. Transport
Problems, 11(3), 77–89.
Macioszek, E. (2018a). The Comparison of Models for Follow-up Headway at Roundabouts. In: Macioszek, E., Sierpiński, G. (eds.) Recent
Advances in Traffic Engineering for Transport Networks and Systems. LNNS 21, 16–26. Springer, Switzerland.
Małecki, K. (2016). The importance of automatic traffic lights time algorithms to reduce the negative impact of transport on the urban
environment. Transportation Research Procedia, 16, 329–342.
Małecki, K. (2017). Graph Cellular Automata with Relation-Based Neighbourhoods of Cells for Complex Systems Modelling: A Case of Traffic
Simulation. Symmetry, 9(12), 322.
Małecki, K. (2017a). The Use of Heterogeneous Cellular Automata to Study the Capacity of the Roundabout. In: Rutkowski L., Korytkowski M.,
Scherer R., Tadeusiewicz R., Zadeh L., Zurada J. (eds) Artificial Intelligence and Soft Computing. ICAISC 2017. LNCS 10246. Springer,
Cham.
Małecki, K. (2018). The Roundabout Micro-simulator Based on the Cellular Automaton Model. In: Sierpiński G. (ed.) Advanced Solutions of
Transport Systems for Growing Mobility. TSTP 2017. Advances in Intelligent Systems and Computing 631. Springer, Cham.
Małecki, K., & Iwan, S. (2012). Development of Cellular Automata for Simulation of the Crossroads Model with a Traffic Detection System. In:
Mikulski J. (ed.) Telematics in the Transport Environment. TST 2012. CiCIS 329. Springer, Berlin, Heidelberg.
Małecki, K., & Wątróbski, J. (2017). Cellular Automaton to Study the Impact of Changes in Traffic Rules in a Roundabout: A Preliminary
Approach, Applied Sciences, 7(7), 742.
Małecki, K., Pietruszka, P. (2018). Comparative Analysis of Chosen Adaptive Traffic Control Algorithms. In: Macioszek E., Sierpiński G. (eds)
Recent Advances in Traffic Engineering for Transport Networks and Systems. TSTP 2017. LNiNS 21. Springer, Cham.
Małecki, K., Pietruszka, P., Iwan, S. (2017). Comparative Analysis of Selected Algorithms in the Process of Optimization of Traffic Lights, In:
Nguyen N., Tojo S., Nguyen L., Trawiński B. (eds) Intelligent Information and Database Systems, LNCS 10192, 497–506.
Mamei, M., Roli, A., & Zambonelli, F. (2005). Emergence and control of macro-spatial structures in perturbed cellular automata, and
implications for pervasive computing systems, IEEE Trans. on Systems, Man, and Cybernetics-Part A: Systems and Humans, 35(3), 337–348.
Missoum, S., Gürdal, Z., Setoodeh, S. (2005). Study of a new local update scheme for cellular automata in structural design. Structural and
multidisciplinary optimization, 29(2), 103–112.
Nagel, K., & Schreckenberg, M. (1992). A cellular automata model for freeway traffic. Journal de Physique I, 2, 2221–2229.
Nagel, K., Wolf, D. E., Wagner P., & Simon P. M. (1998). Two-lane traffic rules for cellular automata: A systematic approach, Physical Review
E, 58(2), 1425–1437.
Nowosielski, A., Frejlichowski, D., Forczmański, P., Gościewska, K., Hofman, R. (2016). Automatic Analysis of Vehicle Trajectory Applied to
Visual Surveillance. Image Processing and Communications Challenges 7, Advances in Intelligent Systems and Computing 389, 89–96.
Rickert, M., Nagel, K., Schreckenberg, M., & Latour, A. (1996). Two lane traffic simulations using cellular automata. Physica A Stat. Mech.
Appl., 231, 534–550.
Schadschneider, A., & Schreckenberg, M. (1997). Traffic flow models with ‘slow-to-start’rules, Ann. Phys., 509(7), 541–551.
Sierpiński, G. (2012). Theoretical Model and Activities to Change the Modal Split of Traffic. In: J. Mikulski (ed.) Telematics in the Transport
Environment, CiCIS 329, 45–51. Springer-Verlag Berlin Heidelberg.
Smith, L., Beckman, R., Baggerly, K., Anson, D., & Williams, M. (1995). Overview of TRANSIMS, the TRansportation ANalysis and
SIMulation System (Combined One-Pagers). Los Alamos National Security: New Mexico, NM, USA.
Sobota, A., Klos, M. J., & Karoń, G. (2017). The Influence of Countdown Timers on the Traffic Safety of Pedestrians and Vehicles at the
Signalized Intersection. In: Sierpiński G. (ed.) Intelligent Transport Systems and Travel Behaviour. Advances in Intelligent Systems and
Computing 505. Springer, Cham.
Touhbi, S., Babram, M. A., Nguyen-Huu, T., Marilleau, N., Hbid, M. L., Cambier, C., & Stinckwich, S. (2017). Adaptive Traffic Signal Control:
Exploring Reward Definition For Reinforcement Learning. Procedia Computer Science, 109, 513–520.
Wang, Y. M., Zhou, L. S., & Lv, Y. B. (2008). Cellular automaton traffic flow model considering flexible safe space for lane-changing. J. Syst.
Simul., 20, 1159–1162.
Wąs, J., Sirakoulis, G. C. (2015). Cellular automata applications for research and industry. Journal of Computational Science, 11, 223–225.
Wolfram, S. (2002). A new kind of science. Champaign: Wolfram media.
Yang, Q., & Koutsopoulos, H. N. (1996). A microscopic traffic simulator for evaluation of dynamic traffic management systems. Transp. Res. C
Emerg. Technol., 4, 113–129.
Zamith, M., Leal-Toledo, R. C. P., Clua, E., & Toledo, E. M. (2015). A new stochastic cellular automata model for traffic flow simulation with
drivers’ behavior prediction. Journal of Computational Science, 9(1), 51–56.
Zhu, H. B., Zhang, N. X., Wu, W. J. (2015). A modified two-lane traffic model considering drivers’ personality, Physica A, 428, 359–367.
Vol. 2 No.1 Januari 2019 Rang Teknik Journal
http://jurnal.umsb.ac.id/index.php/RANGTEKNIKJOURNAL
ANALISIS TRANSPORTASI TERHADAP PENERAPAN ARUS LALU LINTAS

SATU ARAH
ISHAK
Prodi Teknik Sipil UMSB
ishakumsb@gmail.com
Abstrak: Perubahan arus lalulintas satu arah (One Way Street) pada ruas jalan yang diterapkan
Pemerintah Kota Jambi tentunya akan berimbas pada pengguna jalan yang semakin meningkat
karena menghubungkan ke jalan utama. Dengan semakin meningkatnya para pengguna jalan
(kendaraan bermotor) maka akan mempengaruhi kinerja dan tingkat pelayanan jalan tersebut.
Metodologi penelitian yang digunakan dalam penelitian ini memiliki tahapan-tahapan, mulai dari
studi literatur dan pengumpulan data secara primer dan sekunder. Selanjutnya pengumpulan data
hambatan samping dilakukan dengan menghitung dan mencatat jumlah aktivitas samping jalan pada
lokasi penelitian selama periode pengamatan yang dilakukan oleh beberapa surveyor. Hasil
perhitungan kecepatan arus bebas Kol. Amir Hamzah yang menggunakan arus satu arah didapat
sebesar 51 km/. Derajat kejenuhan ( DS) atau tingkat pelayanan jalan satu arah (One Way Street)
Kategori C dan Kategori B, berarti arus lalu lintas masih baik (lancar).
Kata Kunci: Transportasi, arus lalu lintas, satu arah
PENDAHULUAN Dengan skala yang cukup besar, ternyata

Peningkatan pembangunan pada suatu sangat mempengaruhi terhadap pertumbuhan
wilayah dan peningkatan jumlah penduduk pergerakan lalulintas di pusat Kota Jambi
menyebabkan perlu diperhatikan sarana khususnya di ruas Jalan Kol. Amir Hamzah.
maupun prasarana di bidang transportasi Pemerintah Kota Jambi pada tahun 2013
khususnya jalan, Sebagai salah satu urat nadi merubah arah lalulintas yang tadinya Jalan
pembangunan jalan diperlukan perhatian dua arah menjadi jalan satu arah (One Way
secara khusus. Demikian pula pertumbuhan Street)dimulai pukul 06.00-18.00.
penduduk dan kendaraan yang meningkat Perubahan arus lalulintas satu arah (One Way
menyebabkan kemacetan. Hal ini terjadi tidak Street) pada ruas jalan yang diterapkan
ada keseimbangan antara pertumbuhan Pemerintah Kota Jambi tentunya akan
kendaraan dan penduduk dengan penambahan berimbas pada pengguna jalan yang semakin
ruas jalan. Untuk itu perlu dilakukan terobosan meningkat karena menghubungkan ke jalan
atau rekayasa lalulintas untuk meningkatkan utama. Dengan semakin meningkatnya para
kinerja jalan tersebut. pengguna jalan (kendaraan bermotor) maka
Adapun Peraturan Pemerintah akan mempengaruhi kinerja dan tingkat
Republik Indonesia no 32 tahun 2011 tentang pelayanan jalan tersebut.Manual Kapasitas
manajemen dan rekayasa, analisa dampak, Jalan Indonesia (MKJI) tahun 1997 yang
serta manajemen kebutuhan lalulintas. diterbitkan oleh Direktorat Jenderal Bina
Peraturan Pemerintah yang dimaksud adalah Marga Departemen Pekerjaaan Umum
untuk mengoptimalkan penggunaan jaringan Republik Indonesia, merupakan suatu referensi
jalan dan gerakan lalulintas dalam rangka untuk panduan dalam bidang rekayasa
menjamin keamanan, keselamatan, ketertiban lalulintas seperti persimpangan bersinyal,
dan kelancaran lalulintas dan angkutan jalan. persimpangan tidak bersinyal, bagian jalinan
Kota Jambi merupakan salah satu kota yang jalan, jalan perkotaan, jalan luar kota dan jalan
cukup pesat perkembangannya terutama di bebas hambatan.
pusat-pusat kota dengan berbagai aktifitas MKJI disusun berdasarkan hasil studi
seperti perdagangan, jasa, pendidikan, dan penelitian fasilitas jalan perkotaan di 147
perkantoran, rumah sakit dan permukiman. lokasi pada 16 kota dan untuk jalan luar kota
Menurut Badan Pusat Statistik (BPS) Kota dan jalan bebas hambatan pada 128 lokasi di 8
Jambi, Jumlah penduduk Kota Jambi pada provinsi diseluruh Indonesia. Sebelum
tahun 2016 adalah sebesar 576,067 jiwa diterbitkannya MKJI maka telah melakukan
dengan rata-rata kepadatan 3244 jiwa/km. survey pada 275 lokasi di seluruh Indonesia.
ISSN 2599-2081 Fakultas Teknik UMSB 57
EISSN 2599-2090
Penelitian dan studi untuk mendukung kecepatan rata-rata tinggi dan jumlah jalan
terbitnya MKJI dilaksanakan dalam kurun masuk dibatasi secara berdaya guna.
waktu dari tahun 1991 sampai tahun 1996, b) Jalan kolektor merupakan jalan umum yang
yang mana pada setiap lokasi survey diperoleh berfungsi melayani angkutan pengumpul
rekaman video yang menerus dari semua dengan ciri-ciri perjalanan jarak sedang,
gerakan lalulintas pada suatu fasilitas jalan kecepatan rata-rata sedang dan jumlah jalan
dari pagi sampai sore untuk selanjutnya diolah masuk dibatasi.
dan dianalisa. Dengan dukungan dari studi dan c) Jalan lokal merupakan jalan umum yang
survey tersebut maka MKJI menjadi manual berfungsi melayani angkutan setempat
yang benar-benar empiris untuk kondisi di dengan perjalanan jarak dekat, kecepatan
Indonesia. Sebelum diterbitkannya MKJI, para rata-rata rendah dan jumlah jalan masuk
peneliti dan para ahli bidang rekayasa lalu tidak dibatasi.
lintas dalam melaksanakan studi lalulintas d) Jalan lingkungan merupakan jalan umum
selalu mempergunakan panduan atau manual yang berfungsi melayani angkutan
dari berbagai negara Eropa dan Amerika lingkungan dengan ciri-ciri perjalanan jarak
Serikat ataupun Australia yang kondisinya dekat dan kecepatan rata-rata rendah.
tidak sama dengan Indonesia, sehingga hasil Pembagian jalan dalam beberapa kelas
dari studi tersebut kurang dapat mewakilli didasarkan pada kebutuhan transportasi,
keadaan yang sebenarnya. pemilihan moda secara tepat dengan
Pengertian Jalan mempertimbangkan keunggulan karakteristik
Jalan adalah prasarana transportasi masing- masing moda, perkembangan
darat yang meliputi segala bagian jalan, teknologi kendaraan bermotor, muatan sumbu
termasuk bangunan pelengkap dan terberat kendaraan bermotor serta konstruksi
perlengkapannya yang diperuntukkan bagi lalu jalan. Menurut Peraturan Pemerintah No.43
lintas, yang berada pada permukaan tanah, di Tahun 1993 tentang prasarana dan sarana
atas permukaan tanah, di bawah permukaan lalulintas jalan, pada pasal 11 ayat 1 membagi
tanah dan/atau air, serta di atas permukaan air, menjadi 5 (lima) kelas, yaitu:
kecuali jalan kereta api, jalan lori, dan jalan a) Jalan kelas I, yaitu jalan arteri yang dapat
kabel (Peraturan Pemerintah No.34 Tahun dilalui kendaraan bermotor termasuk
2006). Jalan umum adalah jalan yang muatan dengan ukuran lebar tidak melebihi
diperuntukkan bagi lalaulintas umum, jalan 2.500 milimeter, ukuran panjang tidak
khusus adalah jalan yang dibangun oleh melebihi 18.000 milimeter, dan muatan
instansi, badan usaha, perseorangan atau sumbu terberat yang diizinkan lebih besar
kelompok masyarakat untuk kepentingan dari 10 ton.
sendiri. Bagian-bagian jalan meliputi: b) Jalan kelas II, yaitu jalan arteri yang dapat
a) Ruang manfaat jalan meliputi badan jalan, dilalui kendaraan bermotor termasuk
saluran tepi jalan, dan ambang muatan dengan ukuran lebar tidak melebihi
pengamannya. 2.500 milimeter, ukuran panjang tidak
b) Ruang milik jalan meliputi ruang manfaat melebihi 18.000 milimeter, dan muatan
jalan dan sejalur tanah tertentu diluar ruang sumbu terberat yang diizinkan 10 ton.
manfaat jalan. c) Jalan Kelas III A, yaitu jalan arteri atau
c) Ruang pengawasan jalan merupakan ruang jalan kolektor yang dapat dilalui kendaraan
tertentu diluar ruang milik jalan yang ada bermotor termasuk muatan dengan ukuran
dibawah pengawasan penyelenggara jalan. lebar tidak melebihi 2.500 milimeter,
ukuran panjang tidak melebihi 18.000
Klasifikasi Jalan Menurut Fungsinya milimeter, dan muatan sumbu terberat yang
Jalan umum adalah jalan yang diizinkan 8 ton;
diperuntukkan bagi lalulintas umum. Jalan d) Jalan Kelas III B, yaitu jalan kolektor yang
umum menurut fungsinya dikelompokkan ke dapat dilalui kendaraan bermotor termasuk
dalam jalan arteri, jalan kolektor, jalan lokal muatan dengan ukuran lebar tidak melebihi
dan jalan lingkungan. 12.000 milimeter dengan muatan sumbu
a) Jalan arteri merupakan jalan umum yang terberat yang diizinkan 8 ton.
berfungsi melayani angkutan jarak jauh,
58 Fakultas Teknik UMSB ISSN 2599-2081

EISSN 2599-2090
e) Jalan Kelas III C, yaitu jalan lokal yang periode pengamatan yang dilakukan oleh
dapat dilalui kendaraan bermotor termasuk beberapa surveyor.
muatan sumbu dengan ukuran lebar tidak
melebihi 2.100 milimeter dengan muatan HASIL DAN PEMBAHASAN
sumbu terberat yang diizinkan 8 ton. Analisis Volume Lalulintas
Volume lalulintas dijalan Kol. Amir
METODOLOGI PENELITIAN Hamzah menunjukkan volume yang beragam
Metodologi penelitian yang digunakan pada setiap jamnya. Pola pergerakan lalulintas
dalam penelitian ini memiliki tahapan-tahapan, yang terjadi pada jalan ini sangat dipengaruhi
mulai dari studi literatur dan pengumpulan oleh rutinitas pengendara yang melewati dan
data secara primer dan sekunder. Lebih jelas aktivitas disisinya. Berdasarkan hasil survey
dapat diuraikan dalam penjelasan dibawah ini: yang dilakukan selama 2 (dua) hari yaitu pada
1. Studi literatur hari senin dan rabu tanggal 19 dan 21
Studi literatur ini meliputi pengambilan September 2016 selama 12 jam, yang dimulai
teori-teori serta rumus-rumus dari buku, jurnal pukul 06.00 sampai dengan 18.00 WIB,
ilmiah, makalah-makalah seminar yang dengan pencacahan setiap jenis atau
berkaitan dengan penelitian ini. penggolongan jenis kendaraan secara manual
2. Pengumpulan data menurut dengan Manual Kapasitas Jalan
Pengumpulan data terbagi dua kategori Indonesia (MKJI) 1997, yaitu untuk kendaraan
yaitu pengumpulan data sekunder dan data Light Vehcle(LV) atau kendaraan ringan,
primer. Adapun pengumpulan data tersebut Heavy Vehicle (HV) atau kendaraan berat, dan
dijelaskan sebagai berikut: Motor Cycle (MC) atau sepeda motor.
A. Data Sekunder Kemudian dihitung dalam satuan
Data sekunder adalah data yang mobil penumpang (smp), selanjutnya ditentuan
didapat dengan mendatangi instansi nilai faktor ekivalen mobil penumpang (faktor
terkait untuk meminta data-data emp) menurut ketentuan MKJI 1997,yaitu
sekunder yang diperlukan antara lain sebagai berikut:
data kependudukan. Tabel 1. Nilai faktor ekivalen mobil
B. Data primer penumpang (emp)
Data primer adalah data yang didapat Nilai Faktor
No Jenis Kendaraan
dengan pengamatan secara langsung emp
(observasi) dilapangan. Fungsi data
primer dalam penelitian ini adalah 1 Sepeda Motor (SPM) 0,25
untuk mengetahui secara jelas
Kendaraan Ringan
mengenai data volume lalulintas dan 2 1,0
(KR)
hambatan samping. Pengamatan
langsung yang dimaksud untuk Kendaraan Berat
mengetahui komposisi lalu;lintas, 3 1,2
(KB)
komposisi kendaraan dan volume
lalulintas dijalan satu arah (one way Sumber : MKJI 1997
street) Kol. Amir Hamzah. Waktu Data volume lalulintas diambil yang
pelaksanaan perhitungan volume paling besar volumenya pada saat survey yaitu
lalulintas dari pagi hari sampai sore pada hari senin tanggal 19 september 2016.
hari (06.00-18.00) dan dilakukan pada Kemudian berdasarkan volume kendaraan per-
titik yang ditentukan dilakukan jam tersebut, dilakukan perhitungan dengan
surveyor secara bergantian untuk mengalikan volume lalulintas hasil survey
melihat kendaraan angkutan dengan faktor smp untuk mendapatkan volume
penumpang, barang dan sepeda motor. lalulintas dalam smp/jam.
Selanjutnya pengumpulan data
hambatan samping dilakukan dengan
menghitung dan mencatat jumlah aktivitas
samping jalan pada lokasi penelitian selama

EISSN 2599-2090
Tabel 2. Data Volume Lalulintas Jalan Kol. sepanjang 200 meter sesuai dengan ketentuan
Amir Hamzah Manual Kapasitas Jalan Indonesia
(MKJI)1997, yaitu pejalan kaki (Pedestrian),
Arus lalulintas Vol Arus lalulintas
kendaraan/jam ume smp/jam
Vol kendaraan parkir atau berhenti (Parking and
. Stop Vehicles), kendaraan keluar dan masuk
Ken M
Waktu H sm
M H dar C* LV jalur (Exit and Entry Vehicles) dan kendaraan
LV V* p/j
C V aan/ 0.2 *1
jam 5
1.2 am lambat (Slow Vehicles). Kemudian data diolah
06.00- 14 210 35 104
dan cara perhitungannya adalah sebagai
693 2 693 2 berikut:
07.00 08 3 2 7
07.00- 12 183 31
1. Total semua kejadian hambatan
557 2 557 2 877 samping per 1 jam adalah jumlah semua
08.00 75 4 8
kejadian hambatan samping selama 15
08.00- 91 120 22
293 4 293 5 526 menit yang telah dijumlahkan per
09.00 1 8 8
masing-masing faktor kejadian.
Tabel 2. Lanjutan 2. Masing-masing faktor hambatan
samping tersebut kemudian dikalikan
09.00- 76 19
10.00 2
176 7 952
1
176 8 375 dengan faktor bobot kejadian menurut
Manual Kapasitas Jalan Indonesia
10.00- 71 17
11.00 2
274 10 996
8
274 12 464 (MKJI) 1997 adalah sebagai berikut:
Tabel 3. Hasil Perhitungan Hambatan
11.00- 86 119 21
12.00 1
335 2
8 5
335 2 552 Samping Jalan Kol. Amir Hamzah
Frekuensi Frekuensi
12.00- 85 114 21 Kejadian/200/jam Kejadian/200/jam
276 6 276 7 498 Kelas
13.00 8 0 5 S
M To Hamb
13.00- 10 140 27 Waktu S PE PS EE V atan
320 0 320 0 592 PE PS EE tal
14.00 86 6 2 M D* V* V* * Sampi
D V V
V 0,5 1,0 0,7 0 ng
,
4
14.00- 10 144 25 23 29 23 20 48 Sedan
425 3 425 4 683 06.00-07.00 105 5 53 2
15.00 17 5 4 2 3 2 1 8 g
16 24 18 24 13 45 Sedan
15.00- 10
444 2
146 25
444 2 702
07.00-08.00 3 82 1
16.00 23 9 6 4 0 8 0 2 5 g
16.00- 12 161 32 Tabel 3. Lanjutan

324 2 324 2 648
17.00 87 3 2
08.00- S.Ren
20 26 33 5 10 26 23 2 61
17.00- 11 139 28 09.00 dah
323 4 323 5 609
18.00 25 8 1 09.00- 10 Renda
76 21 69 6 38 21 48 2
Sumber : Survey Lapangan Dan Analisis Data, 10.00 9 h
2016 10.00- 10 Renda
64 27 96 12 32 27 45 5
11.00 9 h
Analisis Hambatan Samping Jalan Kol. 11.00- 13 Renda
104 8 92 16 52 8 73 6
12.00 9 h
Amir Hamzah
12.00- S.Ren
Data hambatan samping yang diambil 13.00
42 26 60 7 21 26 29 3 79
dah
dari lokasi penelitian dengan menggunakan 13.00- 11 10 11 21 Renda
surveyor. Pencacahan hambatan samping 54 5 27 72 2
14.00 7 3 7 8 h
dilakukan secara manual yang ditempat pada 14.00- 28 14 28 48 Sedan
197 7 99 99 3
lokasi yang telah ditentukan, yang dimulai 15.00 2 1 2 3 g
pada pukul 06.00-18.00 Wib selama 2 hari 15.00-
158
13 13
4 79
13
91 2
30 Sedan
yang dilaksanakan pada tanggal 19 dan 21 16.00 4 0 4 6 g
September 2016. Data yang diambil data 16.00- 11 Renda
30 60 52 14 15 60 36 6
hambatan samping yang paling besar frekuensi 17.00 7 h
terjadi pada saat survey yaitu pada hari senin 17.00- S.Ren
27 10 18 0 14 10 13 0 37
18.00 dah
tanggal 19 September 2016, dengan interval
waktu 15 menitan dari hambatan samping Sumber: Analisa Perhitungan Data,2016

EISSN 2599-2090
Analisis Kinerja dan Tingkat Pelayanan Tabel 5. Tabel Perbandingan

Analisa perhitungan ini dibedakan Derajat Kejenuhan
menjadi dua, yaitu kinerja dan tingkat Jalan Satu Arah
pelayanan dengan hambatan samping dan Dengan
Hambatan Tanpa Hambatan Samping
tanpa hambatan samping. Kemudian samping
Waktu
menentukan tingkat pelayanan pada jalan Kol. Tingkat Tingkat
Amir Hamzah dengan derajat kejenuhan (DS) Pelayanan Pelayanan
seperti pada perhitungan di bawah ini: Q/C Q/C
06.00-
DS = Q/C 0.35 C 0.33 C
07.00
Keterangan: 07.00-
0.29 C 0.28 C
DS = Derajat kejenuhan 08.00
08.00-
Q = Arus total sesungguhnya (smp/jam) 09.00
0.16 B 0.17 B
C = Kapasitas sesungguhnya (smp/jam) 09.00-
0.12 B 0.12 B
10.00
Tabel 4. Kinerja dan Tingkat Pelayanan
10.00-
Jalan Kol. Amir Hamzah 0.15 B 0.15 B
11.00
Tanpa Hambatan 11.00-
Dengan Hambatan Samping 0.18 B 0.18 B
Samping 12.00
Tingkat Ka Tingkat 12.00-
Waktu Arus Kap. Arus 0.16 B 0.16 B
Pelayanan p. Pelayanan 13.00
13.00-
Q C Q/C Q C Q/C 0.19 B 0.19 B
14.00
06.00- 0.6 14.00-
1885 2548 0.74 D 1885 2921 D 0.23 B 0.22 B
07.00 4 15.00
07.00- 0.5 15.00-
1578 2548 0.62 D 1578 2921 C 0.24 B 0.22 B
08.00 4 16.00
08.00- 0.3 16.00-
947 2921 0.32 C 947 2921 C 0.21 B 0.21 B
09.00 2 17.00
09.00- 0.2 17.00-
675 2859 0.24 B 675 2921 B 0.19 B 0.19 B
10.00 3 18.00
10.00-
835 2859 0.29 C 835 2921
0.2
C Sumber: Analisa Perhitungan
11.00 8
11.00- 0.3
Data,2016
994 2859 0.35 C 994 2921 C
12.00 4
12.00-
896 2859 0.31 C 896 2921
0.3
C SIMPULAN DAN SARAN
13.00 1
13.00- 0.3
Berdasarkan dari pengolahan data dan
1066 2766 0.38 C 1066 2921 C analisa data dapat disimpulkan sebagai
14.00 6
14.00-
1229 2548 0.48 C 1229 2921
0.4
C berikut:
15.00 2
15.00- 0.4
1. Berdasarkan hasil pengamatan dan
1264 2766 0.46 C 1264 2921 C pengukuran dilapangan, ruas jalan Kol.
16.00 3
Tabel 4. Lanjutan Amir Hamzah merupakan jalan perkotaan
16.00- 0.4 0.4 jalan satu arah (One Way Street) dengan
1166 2859 C 1166 2921 C
17.00 1 0 panjang jalan 1.123 m, trotoar tidak ada
17.00- 0.3 0.3 dan kondisi perkerasan tidak cukup baik
1096 2921 C 1096 2921 C
18.00 5 7
Sumber: Analisa Perhitungan Data,2016 2. Hasil perhitungan kecepatan arus bebas
Berdasarkan hasil analisa kinerja dan Kol. Amir Hamzah yang menggunakan
tingkat pelayanan ruas jalan Kol. Amir arus satu arah didapat sebesar 51 km/jam
Hamzah dalam 2 arah lalulintas yang sedangkan arus lalulintas dua arah didapat
mengunakan hambatan samping dan tanpa sebesar 41 km/jam.
hambatan samping didapat pada jam (06.00 3. Dari perhitungan kapasitas ruas jalan
sampai 08.00) berada pada kondisi “D”. Kondisi ini saat di dapat sebesar 1574
Sedangkan pada pukul 09.00 berada pada smp/jam/lajur, lajur efektif di ruas jalan
kondisi “B” serta pada jam10.00 hingga jam Kol.Amir Hamzah mempunyai 2 lajur.
18.00 berada di kondisi “C”.Dibawah ini tabel Untuk kapasitas total adalah sebesar
perbandingan derajat kejenuhan ( DS) antara 2×1574 = 3148 smp/jam. Sedangkan
jalan dalam kondisi saat ini jalan satu arah kapasitas untuk jalan dua arah adalah
(One Way Street) dan jalan dua arah: sebesar 2575 smp/jam.

EISSN 2599-2090
4. Derajat kejenuhan ( DS) atau tingkat

pelayanan jalan satu arah (One Way Street)
Kategori C dan Kategori B, berarti arus lalu
lintas masih baik (lancar).
SARAN
Beberapa hal yang disarankan guna
memperbaiki kondisi kinerja dan tingkat
pelayanan pada ruas jalan Kol. Amir Hamzah
adalah sebagai berikut:
1. Pada kondisi jam puncak pagi hari
hendaknya, dapat diatur lebih baik lagi bagi
kendaraan yang parkir pada badan jalan
agar tidak menggangu kendraan lain yang
lewat. Jika memungkinkan hanya 1 (satu)
sisi saja badan jalan yang diizinkan untuk
parkir sehingga kinerja dan tingkat
pelayanan dapat menjadi lebih baik.
2. Dalam memberikan izin mendirikan
bangunan (IMB), pemilik bangunan seperti
ruko (rumah toko), sekolah dan rumah sakit
mengharuskan para pemilik tersebut
menyediakan fasilitas parkir sesuai
kebutuhan yang telah ditentukan.
3. Jika memungkinkan, dilakukan pelebaran
jalan untuk meningkatkan kinerja dan
tingkat pelayanan sesuai dengan kondisi
lahan di jalan tersebut.
DAFTAR PUSTAKA
Adisasmita Sakti Adji (2011). Jaringan
Transportasi Teori dan Analisis, Graha
Ilmu, Yogyakarta.
Khisty Jotin, C dan Kent Lall, B. (2005).Dasar
Dasar Rekayasa Transportasi. Edisi3,
Penerbit Erlangga, Jakarta.
MKJI (1997). Manual Kapasitas Jalan
Indonesia (MKJI), Direktorat Jenderal
Bina Marga Departemen Pekerjaan
Umum, Jakarta.
Tamin, O. Z. (2000).Perencanaan dan
Pemodelan Transportasi, Penerbit
Institut Teknologi Bandung, Bandung.
Suwardi (2009). Jurnal Teknik Sipil vol.7
No.2, Surakarta
Putranto Surya Leksmono. (2008). Rekayasa
Lalu Lintas, Jakarta.

EISSN 2599-2090
Transportation Research Part B 38 (2004) 441–458
www.elsevier.com/locate/trb
Dynamic one-way traffic control in automated

transportation systems
a,*
Mark Ebben , Durk-Jouke van der Zee b, Matthieu van der Heijden a
a
School of Business, Public Administration and Technology, University of Twente, P.O. Box 217,
7500 AE Enschede, The Netherlands
b
Faculty of Management and Organisation, University of Groningen, P.O. Box 800, 9700 AV Groningen,
The Netherlands
Received 7 September 2000; accepted 23 May 2003
Abstract
In a project on underground freight transportation using Automated Guided Vehicles, single lanes for
traffic in two directions are constructed to reduce infrastructure investment. Intelligent control rules are
required to manage vehicle flows such, that collision is avoided and waiting times are minimised. In contrast
to standard traffic control at intersections, these control rules should take into account significant driving
times along the single lane (in our application up to 8 min). Whereas periodic control rules are often applied
in traffic theory, we focus on adaptive rules such as look-ahead heuristics and dynamic programming al-
gorithms. Numerical experiments show that our control rules reduce waiting times by 10–25% compared to
a straightforward periodic rule. Dynamic programming yields the best results in terms of mean waiting
times.
Ó 2003 Elsevier Ltd. All rights reserved.
Keywords: Traffic control; Simulation; Transportation networks; Heuristics
1. Introduction
In various situations, traffic from two directions shares the same infrastructure. Well-known
examples are roadblocks caused by traffic accidents or road maintenance (Son, 1999). We en-
countered a similar setting in a project on the design of an underground transportation system
near Schiphol Airport in the Netherlands (van der Heijden et al., 2002a). Here Automatic Guided
*
Corresponding author. Tel.: +31-53-4893893; fax: +31-53-4892159.
E-mail address: m.j.r.ebben@utwente.nl (M. Ebben).
0191-2615/$ - see front matter Ó 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0191-2615(03)00075-4
442 M. Ebben et al. / Transportation Research Part B 38 (2004) 441–458
Fig. 1. A two-way tube connecting double tube systems.
Vehicles (AGVs) carry freight through underground tubes between various terminals. To reduce
infrastructure investment, some terminals are connected by a single tube for traffic in both
directions. Intelligent control of the driving direction is required to guarantee acceptable order
throughput times. Although this problem is related to traditional traffic systems, an essential
difference is that the behaviour of AGVs can be directly influenced. Moreover, quite precise in-
formation is available on expected arrival times of AGVs at the two-way track entrance, because
the driving behaviour of AGVs is more predictable than that of passenger cars and because routes
are known in advance. To exploit this information, we develop new control rules, focusing on
delay reduction.
We address a tube allowing for alternating traffic as a two-way tube, see Fig. 1 for an example
with connections to double tube systems at both sides. As is shown in the figure, all AGVs have an
identical length, but some of them may be driving in a convoy with minimal mutual distance.
According to a control rule, AGVs arriving at the two-way tube may either enter immediately or
queue until a signal is given that access is allowed. Each control rule generates the system states,
shown in Table 1, in cyclic order. We use the term green time for the time spent in the states ÔbÕ and
ÔdÕ and the term clearance time for the time spent in the states ÔaÕ and ÔcÕ. The clearance time refers
to the period with a red traffic light at both ends, while AGVs still have to leave the two-way tube.
The cumulative time spent in the four system states makes up the cycle time. The time spent in
each system state depends on the driving time in the two-way tube, the arrival distributions at
both sides and the control rule. The starts of the red light intervals ÔaÕ and ÔcÕ can be considered as
decision variables in the operational control. We assume that at each moment t it is known how
many vehicles are present in the two-way tube and that an end-of-clearance signal is available.
Technologically, this can be accomplished by using induction wiring at both ends of the tube.
Similarly, information on future arrivals can be obtained from induction wiring at a certain
distance from both tube entrances. For example, with passage loops at a reasonable distance from
the entrance and given the constant speed of the vehicles, we can derive the number of vehicles
that will be waiting at the entrance.
The purpose of a control rule is to minimise average vehicle delay. Note that we consider the
system as a discrete process, as opposed to what is often assumed in traffic literature. A
straightforward control rule is First-Come First-Served (FCFS). However, this rule will only lead
Table 1
Description of the system states
State Description
a AGVs from the right have to wait, but AGVs from the left cannot yet enter the tube, because it still
contains AGVs driving from right to left
b AGVs from the left can enter and AGVs from the right have to wait
c AGVs from the left are stopped and AGVs from the right wait until the tube is cleared
d AGVs from the right can enter and AGVs from the left have to wait
M. Ebben et al. / Transportation Research Part B 38 (2004) 441–458 443
to a reasonable performance in the case of very low traffic intensity and relatively short driving
times along a two-way section (for example crossroads). This is not true for the Schiphol system,
where a high traffic intensity and long tube driving times are envisaged (5–10 min). A second
obvious option is switching the driving direction periodically; the system spends a fixed time in
both the system states Ôa&bÕ and Ôc&dÕ. Note that the effective green time is not fixed, because it
depends on the preceding clearing time. AGVs cannot enter the tube if AGVs from the other
directions are still in the tube. The advantage of a periodic control rule lies in its simplicity; it
requires no information on system status. An obvious drawback is that information on AGV
arrivals and queues at both two-way tube entrances is not used. In particular, when the traffic
intensities at both sides of the two-way tube fluctuate in time (as is true in our application), a
simple periodic control rule may lead to excessive waiting times.
A third simple rule is to serve the entire queue in one direction first, and then the entire queue in
the other direction. In a static situation without any information on future arrivals, it can be
argued that this convoy rule leads to an optimal solution. However, a simple example shows that
this approach need not to be optimal in a dynamic context when new AGVs arrive while both
queues are processed. Imagine the following situation at some time t. The last vehicle from the
right has entered the tube 0.6 min ago and 10 AGVs are waiting at the left side. The driving time
through the tube equals 6 min, so the AGVs from the left could start driving at t þ 5:4; the AGVs
enter the tube at a rate of 20 vehicles per minute. Another AGV arrives at the right side of the
tube; no other AGVs arrive. If the latter AGV has to wait, the additional waiting time incurred
equals 5.4 (remaining clearance time when the AGV arrives) + 10/20 (convoy at the left enters the
tube) + 6 (passing time of the convoy) ¼ 11.9 min. On the other hand, if the AGV from the right is
allowed to enter the tube, the additional waiting time incurred equals 0.6 min for each AGV
waiting at the left side, so 10 0:6 ¼ 6 min. This example shows that the convoy rule is not
optimal for this case.
Therefore, we developed adaptive control rules with increasing complexity and information
usage (Sections 3 and 4). We focus on control rules that are suitable for on-line usage: they have to
be efficient in terms of computation time. The potential of the new control rules is demonstrated
by an extensive simulation study in a rolling horizon setting (Sections 5 and 6). We will also
discuss the implications of embedding the two-way tube in a closed transportation network
(Section 7). In all cases, we use periodic control and the convoy rule as benchmarks.
2. Literature review
Two-way track control is related to junction control supported by traffic signals, see Newell
(1988) for an overview. In the early days traffic signals were usually scheduled according to a pre-
determined scheme (cf. Bell, 1992). A periodic control rule is a good example of such a scheme.
Newell (1988) gave a relation between the mean waiting times in a deterministic and a stochastic
setting, assuming that the clearance time is deterministic. He showed that a deterministic
approach suffices if the clearance time goes to infinity, but also that this is not applicable for most
realistic settings. Despite the long clearance times in our application, the difference in delay be-
tween the stochastic and deterministic case is still around 50% (Newell, 1988, Eq. 2.3.12), apart
from the fact that we do not face deterministic clearance times. Note that in the analysis of
intersections the vehicles mostly are regarded as a continuous fluid (cf. Newell, 2002). In our
dynamic control rules we exploit the fact that in reality it is a discrete arrival process. In case
arrival rates are not too high, improvements can be expected from dynamic control rules.
Stochastic models for periodic control have been addressed by several authors. Mung et al.
(1996) derived distributions of queue lengths at fixed time traffic signals. Heidemann (1994) de-
rived analytical results on statistical distributions of queue lengths and delays at traffic signals,
given Poisson arrivals. He compared these results with several approximations (Webster, 1958;
Miller, 1968). Hu et al. (1997) extended HeidemannÕs model to the multi-lane case, where multiple
vehicles may enter the traffic intersection simultaneously. As is argued in van der Heijden et al.
(2002b), these models are not applicable to the two-way tube problem, because the effective green
period and the effective red period are not constant and known. Therefore, van der Heijden et al.
(2002b) developed an approximation for the mean waiting time in periodic systems with random
clearance times. Still, none of the papers above exploit the information on the specific system state
(AGVs in queue and en route to the two-way tube) to reduce average waiting times. This is not
surprising, because the availability of such information is usually limited. As explained in the
introduction, the AGV systems that we have in mind can provide this additional information.
Situations analogous to our two-way tube arise in traffic control, when one lane of a two-lane
highway is closed because of maintenance or construction activities. Ceder and Regueros (1990)
and Ceder (2000) used a periodic model and delay formula based on Webster (1958) as input for
an optimisation model determining the lane closure policy and the lane closure length. Cassidy et
al. (1994) expanded a deterministic equation for the delay using a statistical analysis of empirical
data. An explicit control policy has not been defined. Shibuya et al. (1996) optimised green in-
tervals using a regression equation for the delay. They observed discrepancies with simulated
green intervals and concluded that effective control is very difficult under heavy traffic flows. In
Son (1999), expressions for the mean delay have been derived, based on the assumption that
vehicles arriving within a period H after a queue has entered the two-way track are allowed to
enter the two-way track as well (green time extension). Here the value of H has been estimated
empirically using discrete choice techniques. All these papers do not focus on waiting time
reduction using an adaptive, state dependent control policy.
Finally, we note that problems rather similar to the two-way tube problem have been studied in
the machine scheduling literature, particularly in choosing batch sizes for ovens, see e.g. Uzsoy et
al. (1992, 1994), Webster and Baker (1995), van der Zee et al. (1997) and Potts and Kovalyov
(2000). Similarities with the two-way tube system can be found in the fact that goods (AGVs) have
to be ‘‘batched’’, with the processing time of a batch equal to the driving time on the two-way
track. An essential difference is the fact that processing times are fixed for oven systems, because
they are related to static product and process characteristics. However, the processing time of a
convoy in a two-way tube depends on the time between the first and the last AGV in a convoy.
Furthermore, oven systems set restrictions on the number of products in a batch, but there is no a
priori limit on the number of AGVs in a convoy.
Summarising, the most important differences compared to literature are:
(a) Significant throughput time in the two-way tube (several minutes).

(b) Focus on adaptive, state-dependent control policies, using information on queue length and/
or future AGV arrivals.
(c) Embedding adaptive control rules in a rolling horizon environment.

(d) No maximum convoy (batch) size.
3. Assumptions and control rule concepts
3.1. Assumptions
Our models are based on the following assumptions:
(1) All AGVs are identical.

(2) When driving, the AGVs travel at the same, constant speed.
(3) Queuing AGVs accelerate instantaneously to their normal speed, when activated, and AGVs
can stop instantaneously when arriving at a queue.
(4) AGVs waiting in queue enter the shared lane with a fixed minimum succession time (d).
Instantaneous acceleration and deceleration (assumption 3) is obviously not realistic, but the
impact on the system performance is low if little time is required compared to the driving time in
the two-way tube. Furthermore, it is possible to extend our methods to include the acceleration/
deceleration time. In practice, the succession time d (assumption 4) will be strictly positive because
of the length of an AGV; two vehicles cannot enter a single lane simultaneously. Also, safety
margins can be included to minimise the collision probability, taking into account the decelera-
tion. Note that we do not make an explicit assumption about the arrival processes, both regarding
inter-arrival distribution and lot-size (single or batch arrivals).
In the remainder of this section we characterise the rules for real-time two-way tube control in
terms of information usage and decision structure. Mathematical details are described in Section 4.
3.2. Characterisation of control rules
For the construction of decision rules we chose minimum average waiting time at the two-way
tube as criterion. The waiting time is defined as the time from the arrival of an AGV at the en-
trance of the two-way tube until it enters the two-way tube. Still, the basic ideas presented in this
paper can also be applied to some other criteria, such as minimum lateness. Our decision rules
may use information on the system state at some decision moment t0 . We distinguish three options
for information availability:
(a) No information on AGV queues and arrivals, only average AGV arrival intensities are
known.
(b) Only local information on queue lengths at the two-way tube entrances is available.
(c) Prior information, i.e. both on current queue lengths and on future arrivals within a certain
information horizon H > 0, is available.
At each decision moment the allowed decisions are: ‘‘let the green traffic light remain green until
the next decision moment’’ or ‘‘put the green light on red so that a change of direction is enforced
Table 2
Classification of the control rules
Decision option Information availability
No information Local information Prior information
One change Periodic Control (PCR) Adaptive local control Adaptive look-ahead control
Convoy rule (Conv) (ARloc) (ARla)
Multiple changes Dynamic Programming (DP)
immediately after this decision moment’’. Policies open to the controller are either to take a single
decision on the timing of the next direction change, or to make a schedule of multiple direction
changes within the information horizon H , i.e. all known AGV arrivals are included in the
schedule.
In Table 2 we classify our control rules according to the information availability and the de-
cision options. If no information on the AGVs waiting or arriving is available, a periodic control
rule and a convoy rule are realistic strategies. The convoy rule does not need any parameter setting;
the direction is changed directly after the last vehicle of a convoy has entered the two-way tube.
The fixed time interval P of the periodic control rule can be based on off-line optimisation,
considering the historical and/or expected traffic flows (cf. van der Heijden et al., 2002b). In the
other cases a decision policy has to be constructed.
A simple adaptive local control rule (Section 4.1) uses information on queue lengths only to
make a decision on the next direction change. Multiple changeovers can be determined by a
dynamic programming approach that yields an optimal solution in a static situation. If only local
information on queue lengths is available, obviously the whole queue at the right (left) first passes
the two-way tube if it is decided that the direction is changed to right–left (left–right). Any other
option, for example processing the queue at the right in two separate batches, is inferior in terms
of total waiting time for the vehicles in queue because of the additional clearance time in case of
an additional direction change. For this reason, the dynamic programming approach naturally
leads to batch processing, equivalent to adaptive local control; the entire queue present at the
decision moment enters the two-way tube.
We consider two control rules based on information on future arrivals. First, we develop an
adaptive look-ahead control rule for determining the next direction change (Section 4.2). Second, a
dynamic programming rule is introduced (Section 4.3), which computes an optimal sequence of
direction changes. Because it is well known that an optimal algorithm for a static situation may be
inferior to a simple heuristic in a rolling horizon setting, we analyse both options in the remainder
of this paper. In a dynamic situation, information changes when additional AGVs arrive and/or
when forecasts of future AGV arrivals are improved. As time proceeds, new information becomes
available and rescheduling is necessary. This implies that the rescheduling frequency in a rolling
horizon approach appears as an additional parameter. Regarding decision policies that determine
multiple changeovers, it is natural to only implement the first changeover in the sequence, taking
future direction changes into account. The other decisions will be overruled by information up-
dates.
Finally, we note that we restrict the decision moments to the following list of discrete
events:
(I) the end of state ÔaÕ, where the two-way tube is cleared for traffic from the left;
(II) the arrival of a new AGV at the left or the right end of the two-way tube in state ÔbÕ, with a
green traffic light at the left entrance;
(III) the end of state ÔcÕ, where the two-way tube is cleared for traffic from the right;
(IV) the arrival of a new AGV at the left or the right end of the two-way tube in state ÔdÕ, with a
green traffic light at the right entrance.
Hence, there are no decision moments during clearance periods (states ÔaÕ and ÔcÕ); we do not
reconsider a decision to change the driving direction.
4. Control rules
Here we address the mathematical details of the adaptive control rules as introduced in Sec-
tion 3.2. For simplicity, we first discuss each control rule assuming that the minimum succession
time d ¼ 0. Next, we generalise our results to d > 0. We use the following basic notation:
t0 decision moment
H information horizon, i.e. at a decision moment t0 , AGV arrivals up to t0 þ H are known
Cðt0 Þ the clearance time of the two-way tube at t0 , i.e. the time needed for the last AGV to leave
the two-way tube from time t0 on if no additional AGVs enter the two-way tube
tir ; til the ith arrival at the right (left) side of the two-way tube, where tir ðtil Þ 6 t0 if an AGV has
already arrived
qr ðtÞ; ql ðtÞ the number of AGVs in queue at the right (left) at time t
Note that the clearance time Cðt0 Þ depends on the AGV positions in the tube at time t0 and is in
general not equal to the driving time in the tube T .
4.1. Adaptive local control (ARloc)
Only the queue size at either side of the two-way tube is known. Without loss of generality, we
assume that the current driving direction is from left to right. Recall that we only need to consider
two options (Section 3.2): either the queue at the left passes the two-way tube first or the queue at
the right passes the two-way tube first. Note that in a dynamic situation this is substantially
different from the convoy rule, because of additional decision moments caused by new arrivals
after convoy entrance. Our optimisation criterion is minimum average waiting time. In a dynamic
situation, this means that we should only account for additional waiting times that can still be
influenced by the decisions to be taken. Waiting times already incurred at the decision moment are
not relevant anymore. Therefore, our adaptive control rules minimise a cost function equal to the
total additional waiting times incurred by all AGVs to be scheduled. Still, we will measure the
average delay at the two-way tube when evaluating the control rules in our simulation study
(Section 6).
4.1.1. The case d ¼ 0

A direction change means that the AGVs at the right side will pass the two-way tube first. This
will induce waiting time for the AGVs in queue at the left, being at least equal to the clearance
time (Cðt0 Þ) plus the driving time (T ). So the costs caused by a direction change are
Wchange ¼ ql ðt0 ÞðCðt0 Þ þ T Þ ð1Þ
The alternative is that the direction is not changed immediately, but after the AGVs from the left
have entered the two-way tube. Then, all AGVs in queue at the right (qr ðt0 Þ) incur additional
waiting time, equal to the difference between the clearance time before making a decision (Cðt0 Þ)
and the new clearance time caused by AGVs entering the two-way tube from the left ( ¼ the
driving time T ):
Wnochange ¼ qr ðt0 ÞðT Cðt0 ÞÞ ð2Þ
The direction is changed immediately when Wchange < Wnochange . Otherwise, in a static situation the
direction change is planned after all AGVs in the queue at the left entrance have entered the two-
way tube. In a dynamic situation, it will be natural to make a new decision when the next AGV
arrives, which makes it different from the convoy rule.
4.1.2. The case d > 0

Eqs. (1) and (2) have to be extended to account for the time it takes a convoy to enter the two-
way tube. Except for the first AGV, all AGVs have to remain at a distance d from their prede-
cessor:
Wchange ¼ ql ðt0 ÞðCðt0 Þ þ T þ d maxfqr ðt0 Þ 1; 0gÞ ð3Þ
Wnochange ¼ qr ðt0 ÞðT Cðt0 Þ þ d maxfql ðt0 Þ 1; 0gÞ ð4Þ
4.2. Adaptive look-ahead control (ARla)
Look-ahead control is an extension of local control, taking into account future arrivals within
the information horizon H. Whereas local control leads to a simple choice between two options
(either the convoy from the right or from the left has priority), we can now change the driving
direction at the decision moment or after the ith arrival from the left, see the example in the
introduction.
4.2.1. The case d ¼ 0

The costs related to a direct direction change are a straightforward extension of formula (1):
X
Wchange ¼ minft0 þ T þ Cðt0 Þ tkl ; T þ Cðt0 Þg ð5Þ
tkl 6 t0 þT þCðt0 Þ
Note that this equation reduces to (1) when tkl ¼ t0 . The formula includes the waiting time of
arrivals (tkl ) at the left entrance of the two-way tube up to t0 þ Cðt0 Þ þ T unless, given the
restrictions set on the information horizon H , this information is not available. Note how (5) also
covers those situations in which decision making does not coincide with the arrival of an AGV
(tkl < t0 ).
The alternative is to schedule the direction change after the ith AGV entering the two-way tube.
At that time qr ðtil þ T Þ AGVs are queuing at the right entrance. For each option costs are com-
puted as
X X
Wnochange;i ¼ ðtil þ 2T tkl Þ þ ðtil þ T maxftkr ; t0 þ Cðt0 ÞgÞ ð6Þ
til <tkl <til þ2T tkr <til þT
The first term considers the waiting times of AGVs arriving at the left side of the two-way tube
within the interval ðtil ; til þ 2T Þ, i.e., after i AGVs have entered the two-way tube and before the
scheduled AGVs from the right have passed the two-way tube. The second term of (6) represents
waiting times for AGVs in the queue and arriving at the right side of the two-way tube up to a
time horizon til þ T .
A straightforward comparison between the options is unfortunately not possible, because we
compute waiting times for different schedules. In formula (5), we take into account all waiting
times incurred until t0 þ T þ Cðt0 Þ, whereas in (6) we use all waiting times up to time til þ 2T .
These numbers represent the planning horizon in the various decision options. Obviously, the total
additional waiting time increases with the length of the planning horizon. For fair comparisons,
we weigh the waiting times with the length of the planning horizon, thereby obtaining the ad-
ditional waiting time incurred per time unit. Weighted costs associated with each option are
formulated as

Wchange Wnochange;i
WCchange ¼ and WCnochange ¼ min l ð7Þ
T þ Cðt0 Þ i ti þ 2T t0
The direction is changed immediately if WCchange < WCnochange , otherwise the changeover is planned
after the ith arrival which causes the minimum WCnochange . Note that this planning can be over-
ruled by later decisions.
4.2.2. The case d > 0

Again, the additional waiting time caused by the required distance between AGVs has to be
added:
X
Wchange ¼ minft0 þ T þ Cðt0 Þ þ d ðqr ðt0 þ Cðt0 ÞÞ 1Þ
tkl 6 t0 þT þCðt0 Þþdðqr ðt0 þCðt0 ÞÞ1Þ
tkl ; T þ Cðt0 Þ þ d ðqr ðt0 þ Cðt0 ÞÞ 1Þg ð8Þ

X
Wnochange;i ¼ ðmaxftil ; t0 þ d ði 1Þg
til <tkl <maxftil ;t0 þdði1Þgþ2T þdðqr ðmaxftil ;t0 þdði1ÞgþT Þ1Þ
þ 2T tkl þ d ðqr ðmaxftil ; t0 þ d ði 1Þg þ T Þ 1ÞÞ

X
þ ðmaxftil ; t0 þ d ði 1Þg þ T maxftkr ; t0 þ Cðt0 ÞgÞ ð9Þ
tkr <maxftil ;t0 þdði1ÞgþT
We assume that the number in the queue is greater than 0. Otherwise we have to take the max-
imum of qr ðÞ 1 and 0, as is shown in (3) and (4). The number of options may increase as a result
of the entrance time. The weighted costs associated with each option are formulated as
Wchange
WCchange ¼
T þ Cðt0 Þ þ d ðqr ðt0 þ Cðt0 ÞÞ 1Þ

Wnochange;i
WCnochange ¼ min
i maxftil ; t0 þ d ði 1Þg þ 2T t0 þ d ðqr ðmaxftil ; t0 þ d ði 1Þg þ T Þ 1Þ
ð10Þ
4.3. Dynamic programming (DP)
For those situations where multiple direction changes have to be scheduled, we propose a
Dynamic Programming approach. The scheduling decision involves finding the sequence of
convoys which minimises total additional waiting time. Denoting by NL (NR ) the number of AGVs
from the left (right) within the information horizon at t0 , we can formulate the objective function
as
MINffn ðNL ; NR ; t Þg ð11Þ
n¼L;R
Here fn ði; j; tÞ is defined as the minimum total additional waiting time for all AGVs present at the
two-way tube, or arriving within the horizon H, if at time t already i AGVs from the left and j
AGVs from the right have passed the two-way tube (i ¼ 0; 1; . . . ; NL ; j ¼ 0; 1; . . . ; NR ), and the
last convoy passing the two-way tube came from direction n (n ¼ L; R). Hence t is a point in time
at which all AGVs, known at t0 , have been processed. Note that t is just an auxiliary variable for
the recursion.
4.3.1. The case d ¼ 0

Given the initial conditions at the decision moment (t0 ): fn ð0; 0; t0 Þ ¼ 0, the cost function fn ðÞ
can be formulated, starting with the left side, as
fL ði; j k2 ; t0 Þ ¼ min fc½ði k1 ; j k2 ; tÞ; ði; j k2 ; t0 Þ þ fR ði k1 ; j k2 ; tÞg
k1 ¼1;...;i
fR ði; j; t Þ ¼ min fc½ði; j k2 ; t0 Þ; ði; j; t00 Þ þ fL ði; j k2 ; t0 Þg

00
k2 ¼1;...;j
X
y¼i
0
with c½ði k1 ; j k2 ; tÞ; ði; j k2 ; t Þ ¼ maxft tyl ; 0g ð12Þ
y¼ik1 þ1
X
y¼j
and c½ði; j k2 ; t0 Þ; ði; j; t00 Þ ¼ maxft0 tyr ; 0g
y¼jk2 þ1
0
where t ¼ T þ maxft; til g and t00 ¼ T þ maxft0 ; tjr g
We assume that a convoy contains at least 1 AGV, without affecting optimality, because otherwise
there is no reason to change the direction. The function c½ði k1 ; j k2 ; tÞ; ði; j k2 ; t0 Þ computes
waiting times for a convoy of k1 AGVs that has entered the two-way tube at time t and left it at
time t0 . This time is determined by the question of whether the last AGV arrived during the red
(til 6 t) or green period (til > t). The computation times do not necessarily increase enormously with
more arrivals, because not all discrete points in time have to be computed; the algorithm takes
long time steps of length T . AGVs that wait in queue when the traffic light becomes green will all
enter the two-way tube. It is not optimal to cut a queue into two separate convoys.
4.3.2. The case d > 0

In this case only the cost functions and the decision times change, the recursion remains un-
changed.
X
y¼i
c½ði k1 ; j k2 ; tÞ; ði; j k2 ; t0 Þ ¼ maxft tyl þ d ðy i þ k1 2Þ; 0g
y¼ik1 þ1
X
y¼j
ð13Þ
c½ði; j k2 ; t0 Þ; ði; j; t00 Þ ¼ maxft0 tyr þ d ðy j þ k2 2Þ; 0g
y¼jk2 þ1
where t ¼ T þ maxft þ d ðql ðtÞ 1Þ; til g and t00 ¼ T þ maxft0 þ d ðqr ðt0 Þ 1Þ; tir g
0
In the implementation of the algorithm we assume that all AGVs that are waiting in the queue
when the traffic light switches to green enter the two-way tube in one convoy. A queue is not
separated into two convoys, which might happen for a d > 0. This assumption significantly re-
duces computation times.
5. Design of the simulation study
To demonstrate the potential of the strategies for two-way tube control, we set up an extensive
simulation study. We study the behaviour of a two-way tube in an open system (Section 6) as well
as in a closed system (Section 7). The dynamic behaviour in a closed system is hard to predict,
because convoys created by the two-way tube may cause batch arrivals of the return flow of AGVs
at a later point in time. We use the Schiphol-case to conduct simulation experiments for closed
systems.
The design of the simulation study is based on figures from the Schiphol-case. In Table 3, the
experimental factors and their ranges are shown. Note that these settings with respect to arrival
rates and tube length are not substantially different from Cassidy et al. (1994). For convenience,
the control rules are abbreviated as Conv (Convoy rule), PCR (Periodic Control Rule), ARloc
(Adaptive Rule with local control), ARla (Adaptive Rule with look-ahead), and DP (Dy-
namic Programming). The experimental factors concern the arrival patterns of AGVs (interval,
Table 3
Experimental factors
Factor Range
A. Control rule Conv, PCR, ARloc, ARla, DP
B. Average inter-arrival time (min) 0.25, 0.33, 0.5, 1, 1.5, 2, 4
C. Arrival distribution Poisson, Uniform
D. Lot size 1, Uniform(3, 9)
E. Minimum succession time d (s) 0, 3.5
F. Two-way tube length (m) 1000, 2000, 3000
distribution and lot size) and two-way tube length. The driving speed of the AGVs is constant and
equals 6 m/s. For each control rule average waiting time per AGV has been measured for default
settings. These results are shown in Section 6.1. Next, alternative system configurations were
chosen by changing the value of only one of the factors C–F (Section 6.2). Also, we examine the
sensitivity of our results to two important control parameters, the information horizon and the
planning frequency (Section 6.3).
Note that we determine the fixed time interval P of the periodic control rule by using the
formulas derived in van der Heijden et al. (2002b). This results in change-over times of 5.9, 6.1,
6.3, 6.8, 7.1, 7.3 and 8 min corresponding to average inter-arrival times of 0.25, 0.33, 0.5, 1, 1.5, 2
and 4 min, respectively. For the adaptive look-ahead rule (ARLA ), we chose to consider all options
for changing the driving direction within the interval ½t0 ; t0 þ T . Many choices are possible here;
our choice of the interval was motivated by two arguments. First, considering system charac-
teristics, the driving time in the two-way tube forms a natural horizon. Secondly, as a result of
simulation experiments, we found that the choice of the interval is an acceptable trade-off between
the information requirement and the performance improvement. For the information horizon, we
take as initial setting H ¼ 3T . This period logically follows from the above choice to consider all
possible moments for changing the two-way tube direction up to once the driving time in the two-
way tube (T ). Given a maximum clearance time of T , waiting times for AGVs may be influenced
up to 3T . Recall that we will examine the sensitivity of our results to the value of H in Section 6.3.
In our experiments we adopted the batch means method (cf. Law and Kelton, 2000). Each
batch equals one day. The first batch is discarded to account for any start-up bias. We chose the
number of batches such, that the relative simulation error is 1% at most (within a 95% confidence
interval).
6. Analysis of simulation results
6.1. Default settings
The default settings concern single AGVs arriving according to a Poisson process at a 2000-
meter long two-way tube. Obviously, the average waiting time strongly depends on the AGV
arrival interval and tube length. In order to facilitate the comparison of the results, the perfor-
mance for each setting is indicated in terms of normalised average delay, i.e., the average waiting
time at the two-way tube divided by the driving time in the two-way tube (T ). Since we divide time
by time normalised average delay has no dimension.
Fig. 2 shows that the average waiting time of the convoy rule is equal to the driving time in the
tube, as expected from theory, but only in case of very high arrival rates are the results acceptable.
With very high arrival rates it seems that it is no longer beneficial to wait for additional arrivals,
because this causes too much waiting time to waiting vehicles at the other end of the tube. In this
case a convoy rule might be preferred because of its simplicity. The adaptive rules clearly beat the
PCR-rule by percentages up to 25% at low arrival rates. The differences between the adaptive rules
are smaller. The look-ahead rule performs 2–3% better than the local control rule, while the
dynamic programming rule performs 5–8% better than the look-ahead rule. The outcomes clearly
confirm the general proposition that the more information on future arrivals is included in de-
normalized average delay

1.1
1
0.9 PCR
0.8 Conv
0.7 ARloc
0.6 ARla
0.5 DP
0.4
0.3
0 0.5 1 1.5 2 2.5 3 3.5 4
arrival interval (minutes)
Fig. 2. Normalised average delay for the default settings (simulation results).
cision-making, the better the results. Note that the performance differences presented in Fig. 2
may all be considered significant, given a paired t-test with significance level 95%.
A drawback of a dynamic programming approach might be the computational effort. A series
of experiments indicated that even in the worst case evaluated in the simulation study (see Section
6; average inter-arrival time of 0.25 min), the dynamic programming rule requires less than a
second per changeover using a Pentium III 500 MHz. Although computation times for the dy-
namic programming rules tend to increase exponentially for higher arrival rates, this outcome still
leaves a lot of room for practical application.
6.2. Alternative system configurations
As alternative system configurations, we simulated two-way tubes with the other parameters
values from Table 3. Here we briefly discuss our main findings. Regarding two-way tube length,
we found that this factor hardly influences the relative performance of the control rules. The
influence of the arrival distribution on system performance is studied in two ways. In the first
series of experiments the Poisson process with a negative exponential distribution is replaced by a
more ‘‘regular’’ Uniform distribution. Secondly, we considered compound Poisson arrivals, where
the lot-size of arriving AGVs is uniformly drawn from the interval ½3; 9. Note that we increased
the average inter-arrival times (3, 6, 9 and 12 min) correspondingly.
Two conclusions can be drawn from this series of experiments:
1. The greater the irregularity of the arrival pattern, the better the adaptive control rules perform
in comparison to the periodic control rule. This is due to their responsiveness to system state.
2. More irregular arrival patterns lead to lower normalised average delay. More frequently, large
time gaps occur between AGVs, which are efficiently used by the adaptive rules for changing
the two-way tube direction.
These conclusions are illustrated in Fig. 3. The difference between the performances of the peri-
odic control rule and the dynamic programming rule is much smaller for Uniform arrivals, than it
is for Poisson arrivals, usually less than 6%. We also see that normalised average delay is much
smaller for Poisson arrivals, than it is for Uniform arrivals (cf. DP Poisson and DP Uniform).

1
0.9
0.8
PCR Uniform
0.7 DP Uniform
0.6 PCR Poisson
DP Poisson
0.5
0.4
0.3
0 0.5 1 1.5 2 2.5 3 3.5 4
Fig. 3. Periodic control and dynamic programming for Uniform and Poisson arrivals (simulation results).
1.6
1.4
1.2 PCR
Conv
1
ARloc
0.8
ARla
0.6 DP
0.4
0.2
0 0.5 1 1.5 2 2.5 3 3.5 4
Fig. 4. Intermediate distances and AGV length considered, d ¼ 3:5 s (simulation results).
Similar conclusions were found by Glassey and Weng (1991) in the context of dynamic machine
scheduling.
In the previous experiments we considered the symmetrical case. However, if the arrival rates
from the left and right side differ (asymmetric case), we expect that the performance of the pe-
riodic control rule will be worse. This hypothesis was confirmed in an additional experiment.
Given an average inter-arrival time of 0.5 min from one side and 1 min for the other side, the
relative difference between the periodic control rule and the dynamic programming rule rises to
22% (cf. Fig. 2).
Finally, we examine the impact of the succession time d. Fig. 4 presents results for d ¼ 3:5 s,
which is the actual value in the Schiphol-case. For the periodic rule, we use switching periods of
P ¼ 11 12 ; 9 12 ; 8 14 ; 7 12 ; 7; 6 12 ; and 6 14 min corresponding to k equal to 0.25, 0.33, 0.5, 1, 1.5, 2
and 4 min (cf. van der Heijden et al., 2002b). The remaining settings correspond to the default
settings.
The results in Fig. 4 are similar to those in Fig. 2, although we clearly see the effect of safety
precautions in terms of longer waiting times in the case of high traffic intensities. Again, adaptive
rules perform significantly better than both the periodic rule and the convoy rule. It is remarkable
that most of the performance improvement is already obtained by local control using information
on queue sizes only.
1 1

0.9 0.9
0.8 0.8
DPconvoy(4)
DPconvoy(1)
0.7 0.7 DP(4)
DP(1)
0.6 0.6 ARla(4)
ARla(1)
0.5 0.5
0.4 0.4
1000 2000 3000 4000 5000 6000 7000 8000 1000 2000 3000 4000 5000 6000 7000 8000
information horizon (meter) information horizon (meter)
Fig. 5. Sensitivity for the length of the information horizon and planning frequency for k ¼ 1 (left figure) and k ¼ 4 min
respectively (simulation results).
6.3. Sensitivity for the information horizon and planning frequency
An interesting question from a practical perspective is how the performance of the new control
rules relates to the availability of information on future arrivals. To answer this question, we
study the relationship between the length of the information horizon and system performance. In
this section, we express the information horizon in meters rather than in time, because this nat-
urally relates to facilities for vehicle detection like sensors or inductive loops. The experiment
includes the DP-rule and the ARLA rule. As far as the DP rule is concerned, two scheduling
frequencies are considered: per arriving AGV (DP as considered in Section 4.3) and per convoy
(DPconvoy). The scheduling frequency for the DPconvoy rule is related to a planned direction
change, i.e. no new decision is made before the planned convoy has left the two-way tube. The
two-way tube length is set to 2000 meters, the succession time is set to d ¼ 0 and AGVs arrive
according to a Poisson process with average arrival intervals of 1 and 4 min.
Fig. 5 shows the numerical results, where the average inter-arrival time in minutes is given
between parentheses. We see that the dynamic programming rule (DP) and the look-ahead rule
(ARLA ) are not very sensitive to the length of the information horizon. An information horizon of
3000 meters already gives good results, while more distant information horizons do not lead to
considerable improvement. The fact that performance is most influenced at low arrival rates can
easily be explained by the relative weight of information on arriving AGVs under these circum-
stances. Another effect, which is clearly shown by Fig. 5, is the effect of the planning frequency in a
dynamic context; compare the results for both DP rules. Clearly, a lower planning frequency can
significantly reduce system performance. For longer information horizons, the effect is smaller.
7. A two-way tube in a closed system
7.1. The effects of convoys on the system
Until now we have studied the two-way tube in isolation, as part of an open system. As a
consequence, arrivals were considered independently of each other and an infinite population was
assumed. It is reasonable to assume an open system in the case of a road network or a large
transportation network. In these cases, the convoys that are formed by the two-way tube are
dispersed to different destinations and the population is also large compared to the traffic in the
section under consideration. In a closed system these assumptions can no longer be justified,
making it necessary to study the effects of interaction with other elements in the network (compare
the attention paid in the traffic literature to co-ordinating traffic signals, see e.g. Bell, 1992). For
example, let us consider the situation where the network at one side of the two-way tube is a single
terminal. The convoys that leave the two-way tube drive to the terminal, load/unload, and
probably soon return to the same two-way tube. So the two-way tube will face batch arrivals of
AGVs, where the batch size is heavily influenced by a previous direction change. Decisions on
two-way tube control may therefore have serious impact on future decisions, as well as having a
significant impact on total system performance.
7.2. Numerical results for the Schiphol-case
To check whether our conclusions for two-way tube control in an open system are equally
valid for a closed system, we embedded the rules in a simulation model for the Schiphol system.
The layout that was studied contains one two-way tube (see Fig. 6). At one side of the two-way
tube there is only one terminal. At the other side the two-way tube is connected to several
terminals. The two-way tube has a length of 1500 m and for both sides accurate information on
arrivals is available by recording AGV movements 1500 m before the entrance. The required
distance between two AGVs is 3.5 s. The control rules still aim to minimise the waiting
time locally at the start of the two-way tube, but we measure the total system performance in
terms of the service level, i.e. the percentage of transportation jobs that is handled before the due
time.
The question is whether the local control rules that perform best in minimising the average
waiting time at an isolated two-way tube also realise the best system performance. For PCR
after some initial experiments we chose a switching period of 7 min. The simulation results
confirm the ranking of the control rules with respect to our performance measure, which can be
seen in Table 4.
Fig. 6. Layout of the Schiphol system.

Table 4
Overall service level (% on-time delivery) of the Schiphol system using different control rules
Control rule 180 AGVs 190 AGVs
PCR (7 min) 90.2% 99.7%
ARLOC 92.4% 99.5%
ARLA 93.5% 99.6%
DP 99.1% 100%
8. Conclusions
In this paper we addressed the issue of real-time control of a two-way tube. Several new
control rules were defined as an alternative to the classic periodic control rule. The adaptive
control rules try to use local information on queue lengths and look-ahead information on
future arrivals to improve on system performance. Simulation results indicated that, depending
on the amount of information available, adaptive control rules improve system performance up
to 10% for high arrival rates and up to 25% for low arrival rates. Best performance was realised
by a rule based on dynamic programming. While it is known that a dynamic programming
approach may involve high computational costs, simulation results for a realistic large-scale
transportation network indicated that computation times for the dynamic programming rules
are acceptable for real-time applications. In general, the performance of the rules was not much
influenced by the length of the information horizon. An experiment that included the two-
way tube in a closed network, the Schiphol system, showed that our conclusions still hold
for a closed system. An interesting topic for future research is a more thorough examination
of the interaction effects with terminals or other two-way tubes within an open or closed net-
work.
References
Bell, M.G.H., 1992. Future directions in traffic signal control. Transportation Research A 26, 303–313.
Cassidy, M.J., Son, Y.T., Rosowsky, D.V., 1994. Estimating motorist delay at two-lane highway work zones.
Transportation Research A 28, 433–444.
Ceder, A., Regueros, A., 1990. Traffic control (at alternate one-way sections) during lane closure periods of a two-way
highway. In: Koshi, M. (Ed.), Proceedings of the Eleventh International Symposium on Transportation and Traffic
Theory, pp. 551–568.
Ceder, A., 2000. An application of an optimal traffic control during lane closure periods of a two-lane road. Journal of
Advanced Transportation 34, 173–190.
Glassey, C.R., Weng, W.W., 1991. Dynamic batching heuristics for simultaneous processing. IEEE Transactions on
semiconductor manufacturing 4, 77–82.
Heidemann, D., 1994. Queue length and delay distributions at traffic signals. Transportation Research B 28, 377–389.
Law, A.M., Kelton, W.D., 2000. Simulation Modelling and Analysis, third ed. McGraw-Hill, New York.
Miller, A.J., 1968. Settings for fixed-cycle traffic signals. Operations Research Quarterly 14, 373–386.
Mung, G.K.S., Poon, C.K.A., Lam, W.H.K., 1996. Distributions of queue lengths at fixed time traffic signals.
Transportation Research B 30, 421–439.
Newell, G.F., 1988. Theory of highway traffic signals. Report no. UCB-ITS-CN-89-1, Institute for Transportation
Studies, University of California.
Newell, G.F., 2002. The oversaturated isolated intersection. In: Transportation and Traffic Theory in the 21st Century:
Proceedings of the 15th International Symposium on Transportation and Traffic Theory, Adelaide, Australia, pp.
119–135.
Potts, C.N., Kovalyov, M.Y., 2000. Scheduling with batching: a review. European Journal of Operational Research
120, 228–249.
Shibuya, S., Nakatsui, T., Fujiwara, T., Matsuyama, E., 1996. Traffic control at flagger-operated work zones on two-
lane roads. Transportation Research Record 1529, 3–9.
Son, Y.T., 1999. Queueing delay models for two-lane highway work zones. Transportation Research B 33, 459–471.
Uzsoy, R., Lee, C.Y., Martin-Vega, L.A., 1992. A review of production planning and scheduling models in the
semiconductor industry, Part I: System characteristics, performance evaluation and production planning. IIE
Transactions 24, 47–60.
Uzsoy, R., Lee, C.Y., Martin-Vega, L.A., 1994. A review of production planning and scheduling models in the
semiconductor industry, Part II: Shop-floor control. IIE Transactions 26, 44–55.
van der Heijden, M.C., van Harten, A., Ebben, M.J.R., Saanen, Y., Valentin, E., Verbraeck, A., 2002a. Using
simulation to design an automated underground system for transporting freight around Schiphol Airport. Interfaces
32, 1–19.
van der Heijden, M.C., van Harten, A., Ebben, M.J.R., 2002b. Waiting times at periodically switched one-way traffic
lanes: a periodic two-queue polling system with random setup times. Probability in the Engineering and
Informational Sciences 15, 495–517.
van der Zee, D.J., van Harten, A., Schuur, P.C., 1997. Dynamic job assignment heuristics for multi-server batch
operations––a cost based approach. International Journal of Production Research 35, 3063–3093.
Webster, F.W., 1958. Traffic signal settings. Road Research Technical paper 39, Her MajestyÕs Stationary Office,
London.
Webster, S., Baker, K.R., 1995. Scheduling groups of jobs on a single machine. Operations Research 43, 692–703.
JURNAL KARYA TEKNIK SIPIL, Volume 3, Nomor 1, Tahun 2014, Halaman 142
JURNAL KARYA TEKNIK SIPIL, Volume 3, Nomor 1, Tahun 2014, Halaman 142 ± 153
Online di: http://ejournal-s1.undip.ac.id/index.php/jkts
PENERAPAN MANAJEMEN LALU LINTAS SATU ARAH

PADA RUAS JALAN SULTAN AGUNG ± SISINGAMANGARAJA ±
DR.WAHIDIN KOTA SEMARANG UNTUK PEMERATAAN SEBARAN BEBAN
LALU LINTAS
Ramadhania Pramanasari, Nurul Qomariyah, Djoko Purwanto*), Epf. Eko Yulipriyono*)
Jurusan Teknik Sipil, Fakultas Teknik, Universitas Diponegoro

Jl. Prof Soedarto, Tembalang, Semarang. 50239, Telp.: (024)7474770, Fax.: (024)7460060
ABSTRAK
Semarang merupakan ibukota Provinsi Jawa Tengah dengan tingkat pertumbuhan dan
mobilisasi penduduk yang tinggi. Hal ini menyebabkan pertumbuhan kendaraan di Kota
Semarang yang terus meningkat yang berdampak pada penumpukan kendaraan di suatu
ruas jalan tertentu seperti pada ruas Jalan Dr. Wahidin dan Jalan Sultan Agung. Hal ini
menandakan bahwa pada ruas jalan terjadi kepadatan lalu lintas yang tinggi sehingga
tidak menutup kemungkinan di tahun mendatang peningkatan dapat terjadi kembali.
Penelitian ini bertujuan untuk mengidentifikasi kinerja ruas jalan dan simpang serta
mencari solusi penyebaran lalu lintas yang efisien dengan merencanakan dua skenario
manajemen lalu lintas sistem satu arah. Penelitian ini didahului dengan melakukan kajian
pustaka dan survei pendahuluan yang kemudian dilanjutkan dengan pengumpulan data
melalui traffic counting, observasi, dan pengukuran terkait. Analisis kinerja ruas dan
simpang pada kondisi eksisting dan skenario rencana menggunakan MKJI 1997. Dari
hasil perbandingan kinerja antara kondisi eksisting dengan dua skenario yaitu skenario 1
(ruas Jl.Dr.Wahidin, Jl.Sisingamangaraja, dan Jl.Sultan Agung diberlakukan sistem satu
arah) dan skenario 2 (ruas Jl.Dr.Wahidin, Jl.Sisingamangaraja, dan Jl.Sultan Agung
diberlakukan sistem satu arah dengan contra flow pada ruas Jl. Dr.Wahidin dan Jl.Sultan
Agung) didapatkan bahwa skenario terbaik adalah skenario 1. Skenario 1 dapat
menjadikan nilai DS pada Jalan Sultan Agung menjadi menurun sebesar 0,21-0,33 dan
Jalan Dr.Wahidin sebesar 0,33-0,48. Serta setelah dibandingkan dengan metode scoring,
skenario 1 memperoleh skor tertinggi untuk ruas jalan dan simpang. Oleh karena itu,
skenario 1 dapat direkomendasikan.
kata kunci : kepadatan lalu lintas, sistem satu arah, kinerja lalu lintas
ABSTRACT
Semarang is a capital city of Central Java with high rate of civil growth & mobility. This
caused growth of vehicles in Semarang city which keep increasing and has impact in the
YHKLFOH¶V FXPXODWLRQ RQ RQH VHJPHQW RI FHUWDLQ URDG VXFK DV 'U :DKLGLQ VWUHHW DQG
*)
Penulis Penanggung Jawab
142
Sultan Agung street . It indicates that on the segment of the road occured a high traffic
density that there is a possibility in the upcoming years an increase will happen again. The
SXUSRVH RI WKH UHVHDUFK LV WR LGHQWLI\ WKH SHUIRUPDQFH RI WKH URDG¶V VHJPHQW DQG MXQFWLRQ
and to find the solution of efficient traffic spread with planned two one way traffic
management scenarios. This research begins with literature study and preliminary survey
which continued with collecting data through traffic counting, observation, and related
PHDVXUHPHQWV 3HUIRUPDQFH DQDO\VLV RI URDG¶V VHJPHQWV DQG MXQFWLRQV LQ existing
condition and planned scenario using MKJI 1997. As the result of analysis comparation
between existing condition with 2 scenarios which the first scenario (segment of Dr.
Wahidin street, Sisingamangaraja street, and Sultan Agung street enforced one way
system) and the second scenario (segment of Dr. Wahidin street, Sisingamangaraja street,
and Sultan Agung street enforced one way system with contra flow on segment of Dr.
Wahidin street and Sultan Agung street) gained that the best scenario is the first scenario.
First scenario can decrease the DS of Sultan Agung Street become about 0,21-0,33 and
Dr.Wahidin Street about 0,33-0,48. Also after being compared using scoring method, the
ILUVW VFHQDULR JRW KLJKHVW VFRUH IRU URDG¶V VHJPHQW DQG MXQFWLRQV 6o that, scenario 1 can
be recommended.
keywords: traffic density, one way system,traffic performance
PENDAHULUAN
Latar Belakang
Indonesia merupakan negara berkembang yang memiliki pertumbuhan penduduk yang

tinggi, yaitu mencapai 1,5% per tahun. Hal ini berimbas kepada meningkatnya kebutuhan
penduduk akan pendidikan, ekonomi, dan lain-lain yang mendorong meningkatnya
mobilisasi penduduk. Mobilisasi penduduk yang semakin tinggi ini akhirnya berimbas
kepada meningkatnya jumlah kendaraan.
Semarang sebagai ibukota Provinsi Jawa Tengah dengan pertumbuhan penduduknya

sekitar 2,09% mengakibatkan tingkat mobilisasi barang dan jasa meningkat begitu
signifikan dari tahun ke tahun. Kota Semarang dengan luas wilayah 3.723,67 km2 dan
jumlah penduduk mencapai 1.527.433 jiwa pada tahun 2010 hanya memiliki panjang jalan
2.786,28 km untuk jumlah kendaraan bermotor mencapai 167.159 unit (Sumber : BPS
Provinsi Jawa Tengah 2010)
Pertumbuhan kendaraan yang semakin tinggi ini, tidak diimbangi dengan pertumbuhan
jalan yang ada. Sehingga hal ini menyebabkan terjadinya penumpukan suatu kendaraan
disuatu ruas jalan tertentu. Penumpukan kendaraan ini biasa terjadi pada jam-jam puncak
pagi, siang, dan sore hari. Salah satu kemacetan yang sering terjadi pada jam-jam puncak
adalah pada ruas Jalan Teuku Umar yaitu pada ruas jalan yang menuju ke Jatingaleh. Hal
ini dikarenakan adanya penumpukan lalu lintas dari dua ruas jalan, yaitu Jalan Sultan
Agung dan Jalan Dr. Wahidin yang mengakibatkan kendaraan dari arah Jatingaleh menuju
ke Jalan Dr. Wahidin menjadi terhambat karena terhalang oleh kendaraan yang menumpuk
di tengah simpang.
143
Untuk mengatasi hal tersebut, perlu adanya manajemen dan rekayasa lalu lintas yang
dinamis diantaranya adalah penambahan ruas jalan atau manajemen pengaturan arus.
Melihat dari banyak sisi, penambahan ruas jalan menjadi solusi yang kurang efektif dan
efisien dikarenakan keterbatasan lahan serta besarnya biaya yang harus dikeluarkan.
Dengan demikian, manajemen pengaturan arus bisa menjadi solusi efektif dan efisien
karena teknis yang dilakukan hanya penataan dan pengaturan ruas jalan serta simpang yang
berpotensi menimbulkan kemacetan. Bentuk manajemen pengaturan arus yang dapat
diterapkan adalah manajemen lalu lintas sistem satu arah.
Maksud dan Tujuan
Adapun maksud dari penerapan manajemen lalu lintas sistem satu arah, antara lain:
1. Mengidentifikasi kinerja ruas jalan tersebut;
2. Mencari cara/skenario penyaluran lalu lintas yang efisien dengan menerapkan
manajemen lalu lintas sistem satu arah.
Tujuan dari manajemen lalu lintas sistem satu arah ini, yaitu:
1. Melakukan persebaran lalu lintas berbasis pemerataan sehingga pada jalan-jalan kota
yang awalnya sepi dapat menjadi ramai dan yang awalnya ramai menjadi berkurang
tingkat keramaiannya;
2. Menghasilkan kondisi lalu lintas yang tertib, lancar, aman, nyaman dan efisien
karena berkurangnya waktu tundaan pada simpang sehingga antrian menjadi
berkurang;
3. Memberikan masukan kepada Pemerintah Kota Semarang mengenai penataan sistem
transportasi kota.
Lokasi Studi
Lokasi studi yang ditinjau pada Tugas Akhir ini adalah Jalan Sultan Agung, Jalan
Dr.Wahidin, dan Jalan Sisingamangaraja seperti yang ditentukan pada Gambar 1.
Jl. Sisingamangaraja
Jl. Sultan Agung seksi

2
Jl. Sultan Agung seksi Jl. Dr. Wahidin

1
Jl. Teuku Umar
Gambar 1. Lokasi studi
144
KAJIAN PUSTAKA
Manajemen lalu lintas satu arah adalah suatu pola lalu lintas dengan merubah jalan dua
arah menjadi jalan satu arah yang berfungsi untuk meningkatkan kapasitas jalan dan
persimpangan. Sistem ini biasa diterapkan di wilayah perkotaan guna meningkatkan
keselamatan dan kelancaran lalu lintas.
Kelebihan sistem satu arah, diantaranya:

1. Mengurangi jumlah konflik di persimpangan serta memudahkan pengaturan
koordinasi sinyal lampu lalu lintas;
2. Mengurangi kecelakaan lalu lintas, walaupun demikian fasilitas menjadi bertambah
mengingat kecepatan kendaraan meningkat;
3. Meningkatan kecepatan rata-rata kendaraan pada sistem jaringan jalan, walaupun
demikian tidak berarti mempercepat waktu perjalanan;
4. Memungkinkan terjadinya peningkatan ekonomi/pendapatan wilayah, yang
semulanya adalah kawasan yang tenang menjadi ramai.
Kekurangan sistem satu arah, diantaranya:

1. Dapat menyebabkan waktu perjalanan menjadi lebih lama karena harus berputar;
2. Memungkinkan fasilitas bertambah akibat kecepatan kendaraan menjadi lebih tinggi;
3. Menyulitkan penyeberang jalan apabila tidak diberikan tempat penyeberangan
khusus;
4. Menyulitkan angkutan umum apabila tidak disediakan lajur khusus yang berlawanan
arus;
5. Menyulitkan masyarakat yang tidak terbiasa berpergian ke daerah tersebut karena
rute menjadi berputar-putar.
Analisis kinerja hanya terbatas pada ruas jalan dan simpang bersinyal dengan
menggunakan acuan Manual Kapasitas Jalan Indonesia (MKJI) 1997. Variabel yang
menjadi parameter penentu dari kinerja ruas jalan adalah nilai DS, kecepatan penyesuaian
(VLV), dan waktu tempuh (TT), sedangkan pada simpang bersinyal parameter penentunya
adalah kendaraan terhenti rata-rata dan tundaan simpang rata-rata.
METODOLOGI
Secara garis besar, metodologi yang digunakan dalam menyelesaikan penataan lalu lintas
kali ini adalah:
1. Tahap persiapan, berupa studi kepustakaan mengenai hal-hal yang berhubungan
dengan kapasitas dan kinerja jalan dan simpang yang dapat diperoleh dari berbagai
literatur dan internet;
2. Tahap pengumpulan data, dimana data diperoleh dengan survei dan observasi
lapangan berupa volume kendaraan, geometrik jalan, kondisi lingkungan, jumlah
fase, waktu sinyal, dan panjang antrian pada setiap simpang;
3. Tahap analisis data pada setiap survei yang didapat di lapangan. Dari analisis ini,
dapat langsung diperoleh kondisi ruas jalan dan simpang. Dari analisis ini juga akan
didapatkan kinerja ruas jalan dan simpang pada kondisi eksisting;
4. Penataan lalu lintas baru didasarkan pada kondisi terjenuh pada saat eksisting;
145
5. Penataan lalu lintas dilakukan dengan memperhatikan teori kapasitas dan kinerja lalu
lintas pada MKJI 1997. Pengaturan cycle time pada simpang, pemilihan rute baru
yang dapat memberikan peningkatan kinerja ruas jalan dan simpang menjadi lebih
baik.
Untuk melihat bagan alir dari proses pengerjaan tugas akhir, maka tahapan dijelaskan pada
Gambar 2.
PENYAJIAN DAN ANALISIS DATA
Penataan lalu lintas dilakukan dengan membuat dua skenario rencana yang kemudian
dibandingkan dengan kondisi eksistingnya. Pembuatan dua skenario ini bertujuan untuk
mendapatkan skenario optimal ketika sistem satu arah diberlakukan. Untuk mengetahui
perbandingan pola pergerakan antara kondisi eksisting, skenario 1, dan skenario 2
dijelaskan pada Gambar 3.
Berdasarkan pola pergerakan lalu lintas yang ditampilkan pada Gambar 3 ditentukan
bahwa pada skenario 1 diberlakukan sistem satu arah pada Jalan Sultan Agung, Jalan
Sisingamangaraja, dan Jalan Dr. Wahidin, sedangkan pada skenario 2 diberlakukan sistem
satu arah pada Jalan Sultan Agung, Jalan Sisingamangaraja, dan Jalan Dr. Wahidin dengan
contra flow pada Jalan Sultan Agung dan Dr. Wahidin.
Kinerja Ruas Jalan
Berdasarkan hasil analisis yang telah dilakukan diperoleh hasil mengenai masing-masing
kondisi sesuai dengan ketetapan parameter kinerja ruas jalan yang telah ditentukan. Hasil
analisis dari ketiga kondisi tersebut ditampilkan dalam Tabel 1.
Tabel 1. Perbandingan kinerja ruas jalan pada kondisi eksisting, skenario 1, dan skenario 2
Waktu Tempuh
Derajat Kejenuhan Kecepatan Penyesuaian (VLV)
(TT)
Jam (DS) (km/jam)
Ruas Jalan (detik)
Puncak
Skenario Skenario Skenario Skenario Skenario Skenario
Eksisting Eksisting Eksisting
1 2 1 2 1 2
Pagi 0,24 0,70 0,68 59,78 49,5 54,9 45,98 55,3 55,3
Jl. Siang 0,29 0,77 0,75 59,78 49 54,9 46,37 55,8 57,0
Sisingamangaraja
Sore 0,19 0,78 0,75 59,78 48 54,9 46,77 57,0 57,0
Pagi 0,70 0,67 0,64 43 44 48 29,30 28,6 26,0
Jl. Sultan Agung
Siang 0,62 0,69 0,67 46 43 48 27,39 29,3 28,64
seksi 1
Sore 0,93 0,66 0,64 35 45 48,5 36,00 28,0 26,0
Pagi 0,9 0,69 0,66 34 43 48 52,94 41,9 37,5
Jl. Sultan Agung
Siang 0,64 0,71 0,69 42 42 42 42,86 42,9 42,9
seksi 2
Sore 1,01 0,68 0,66 28 43,5 48 64,29 41,4 37,5
Pagi 0,74 0,46 1,53 35 65 28 123,43 66,5 154,3
Jl. Dr.Wahidin
Siang 0,82 0,49 1,60 34 65 28 127,06 67,0 154,3
seksi 1
Sore 0,96 0,48 1,60 29 64 28 148,97 67,5 154,3
= Peningkatan kinerja = Penurunan kinerja
146
START
IDENTIFIKASI MASALAH
STUDI PUSTAKA
INVENTARISASI KEBUTUHAN DATA
SURVEI PENDAHULUAN
PENGUMPULAN DATA
DATA PRIMER DATA SEKUNDER
1. Geometrik jalan 1. Peta Semarang

- Ruas jalan 2. Geometrik jaringan jalan
- Simpang 3. Data statistik jumlah penduduk
2. Kondisi arus lalu lintas kota Semarang
- Volume kendaraan
3. Pengaturan sinyal eksisting
- Fase sinyal
- Waktu antar hijau dan waktu hilang
4. Hambatan samping
5. Kinerja lainnya
DATA CUKUP? TIDAK
YA
ANALISA
KINERJA KONDISI RUAS JALAN DAN KINERJA LALU LINTAS SISTEM SATU
SIMPANG EKSISTING ARAH (ONE WAY)
SKENARIO 1 SKENARIO 2
(tanpa contra flow) (dengan contra flow)
PERBANDINGAN SISTEM SATU ARAH

(ONE WAY) DENGAN KONDISI EKSISTING
PEMBAHASAN HASIL ANALISA
KESIMPULAN DAN SARAN
FINISH
Gambar 2. Diagram Metodologi Penelitian
147
Kondisi Eksisting
Skenario 1 Skenario 2
Gambar 3. Pola pergerakan lalu lintas
Untuk mengetahui keunggulan dari kedua skenario rencana yang telah dibuat, maka perlu
dilakukan perbandingan dengan metode scoring, tujuannya untuk mempermudah dalam
penetapan skenario terpilih. Sistem penilaian menggunakan skala tertentu sesuai dengan
kriteria yang telah ditetapkan. Parameter yang dipergunakan adalah kecepatan rencana
saja. Hal tersebut dikarenakan, nilai dari kecepatan rencana berbanding lurus terhadap nilai
DS dan waktu tempuh.
Hasil scoring ruas jalan antar alternatif skenario, ditampilkan pada Tabel 3 (jam puncak
pagi), Tabel 4 (jam puncak siang), dan Tabel 5 (jam puncak sore), serta kriteria angka
penilaian ditampilkan pada Tabel 2
Tabel 2. Kriteria angka penilaian untuk kinerja ruas jalan
Kecepatan Rencana
Skor
(km/jam)
0 - 29 1 (rendah)
30 - 59 2 (sedang)
> 60 3 (tinggi)
148
Tabel 3. Scoring ruas jalan pada jam puncak pagi
Kecepatan Penyesuaian (VLV)

Ruas Jalan ± Arah (km/jam)
Eksisting Skor Skenario 1 Skor Skenario 2 Skor
Jl.Wahidin 1 35 2 65 3 28 1
Jl.Sisingamangaraja 58 2 49,5 2 49,5 2
Jl.Sultan Agung Seksi 1 43 2 44 2 48,5 2
Jl.Sultan Agung Seksi 2 34 2 43 2 48 2
JUMLAH SKOR 8 9 7
Tabel 4. Scoring ruas jalan pada jam puncak siang

Jl.Wahidin 1 34 2 65 3 28 1
Jl.Sisingamangaraja 58 2 49 2 48 2
JUMLAH SKOR 8 9 7
Tabel 5. Scoring ruas jalan pada jam puncak sore

Jl.Wahidin 1 29 1 64 3 28 1
Jl.Sisingamangaraja 59 2 48 2 48 2
Jl.Sultan Agung Seksi 1 35 2 45 2 48,5 2
Jl.Sultan Agung Seksi 2 28 1 43,5 2 48 2
JUMLAH SKOR 6 9 7
Berdasarkan hasil perhitungan di atas, diperoleh Jumlah skor dari masing-masing kondisi,
yaitu:
a. Kondisi eksisting = (8+8+6) = 22
b. Skenario 1 = (9+9+9) = 27 (hasil terbaik)
c. Skenario 2 = (7+7+7) = 21
Oleh sebab itu, sesuai dengan hasil di atas dapat dikatakan bahwa skenario 1 menunjukkan
hasil yang lebih baik.
Kinerja Simpang Bersinyal
Dalam melakukan analisis kinerja simpang, parameter yang menjadi tolak ukur dalam
penilaiannya sesuai dengan MKJI 1997 adalah tundaan simpang rata-rata. Dalam analisis
ini, wilayah persimpangan yang menjadi studi kajian kondisi eksisting dan skenario 2
adalah simpang Sutan Agung - Dr. Wahidin, Sultan Agung - Sisingamangaraja, dan
149
Sisingamangaraja-Dr. Wahidin. Sedangkan pada skenario 1 tidak dilakukan analisis

simpang dikarenakan titik konflik pada daerah tersebut sudah tidak terjadi sehingga
simpang ini tidak difungsikan. Pada Tabel 7, Tabel 8, dan Tabel 9 ditampilkan hasil
analisis simpang bersinyal dari ketiga kondisi tersebut, serta kriteria angka penilaian untuk
kinerja simpang bersinyal ditampilkan pada Tabel 6.
Tabel 6. Kriteria angka penilaian untuk kinerja simpang bersinyal

Persen Penurunan
Skor
(%)
0 - 25 1 (rendah)
26 - 50 2 (cukup)
51 - 75 3 (sedang)
> 76 4 (tinggi)
Tabel 7. Scoring Simpang Sultan Agung- Sisingamangaraja, Simpang Sultan Agung ±

Dr.Wahidin, dan Simpang Sisingamangaraja ± Dr.Wahidin pada jam puncak pagi
Kondisi Persen Penurunan
Simpang Parameter Skenario Skenario Skenario Skenario
Eksisting Skor Skor
1 2 1 2
Simpang Tundaan
Sultan Agung ± simpang rata-
1050 0 6,448 100 % 4 99,39% 4
Sisingamangaraj rata (det/smp)
a
Simpang Tundaan
Sultan Agung ± simpang rata- 59 0 6,98 100% 4 88,16% 4
Dr.Wahidin rata (det/smp)
Simpang Tundaan
Sisingamangaraj simpang rata- 4,87 0 8 100% 4 -64,27% 0
a ± Dr.Wahidin rata (det/smp)
JUMLAH SKOR 12 8

Dr.Wahidin, dan Simpang Sisingamangaraja±Dr.Wahidin pada jam puncak siang
Eksisting Skor Skor
1 2 1 2
Simpang Tundaan
1415 0 6,47 100% 4 99,54% 4
a
Simpang Tundaan
Tundaan
Simpang simpang rata-
Sisingamangaraj rata (det/smp) 12,47 0 8 100% 4 35,85% 2
a ± Dr.Wahidin
JUMLAH SKOR 12 9
150

Dr.Wahidin, dan Simpang Sisingamangaraja ± Dr.Wahidin pada jam puncak sore

Eksisting Skor Skor
1 2 1 2
Simpang Tundaan
1404 0 6,612 100% 4 99,52% 4
a
Simpang Tundaan
Simpang Tundaan
Sisingamangaraj simpang rata- 6,55 0 6 100% 4 8,39% 1
a ± Dr.Wahidin rata (det/smp)
JUMLAH SKOR 12 9
Berdasarkan hasil perhitungan di atas, diperoleh jumlah skor dari masing-masing kondisi,
yaitu:
a. Skenario 1 = (12+12+12)
= 36 (hasil terbaik)
b. Skenario 2 = (8+9+9)
= 26
Oleh sebab itu, sesuai dengan hasil diatas dapat dikatakan bahwa skenario 1 menunjukkan
hasil yang paling baik.
Di dalam penerapan sistem satu arah pada skenario 1 dapat dilihat bahwa kinerja simpang
lebih baik daripada kondisi eksisting maupun skenario 2 karena tidak terjadi konflik antar
pendekat pada simpang sehingga membuat para pengendara tidak mengalami tundaan.
PENUTUP
Kesimpulan
Berdasarkan analisis data dan pemecahan masalah yang telah dilakukan, maka dapat
diambil beberapa kesimpulan sebagai berikut:
1. Dari keempat ruas jalan eksisting yang memiliki nilai DS > 0,75 adalah JL. Dr.
Wahidin pada jam puncak siang dan sore, JL. Sultan Agung seksi 1 jam puncak sore,
JL. Sultan Agung seksi 2 jam puncak pagi dan sore. Sedangkan pada ruas JL.
Sisingamangaraja nilai DS sangat kecil, hal ini berarti terjadi ketidakmerataan beban
lalu lintas.
2. Dari simpang bersinyal eksisting yang memiliki nilai DS > 0,75 lebih banyak
terdapat pada Simpang Dr. Wahidin-Sisingamangaraja dan Simpang Sultan Agung ±
Sisingamangaraja baik jam puncak pagi, siang, maupun sore.
3. Penerapan skenario 1 (tanpa contra flow) berhasil menurunkan nilai DS pada lima
ruas jalan, yaitu Jl. Sultan Agung Seksi 1 jam puncak sore dari 0,93 menjadi 0,66 ,
ruas Jl. Sultan Agung Seksi 2 jam puncak pagi dari 0,90 menjadi 0,69, ruas Jl.Sultan
151
Agung Seksi 2 jam puncak sore dari 1,01 menjadi 0,68, ruas Jl.Dr. Wahidin Seksi 1
jam puncak siang dari 0,82 menjadi 0,49, dan ruas Jl.Dr.Wahidin Seksi 1 jam puncak
sore dari 0,96 menjadi 0,48. Sedangkan pada skenario 2 (dengan contra flow) hanya
berhasil menurunkan nilai DS pada tiga ruas jalan saja, yaitu ruas Jl. Sultan Agung
Seksi 1 jam puncak sore dari 0,93 menjadi 0,64, ruas Jl.Sultan Agung Seksi 2 jam
puncak pagi dari 0,90 menjadi 0,69, dan ruas Jl.Sultan Agung Seksi 2 jam puncak
sore dari 1,01 menjadi 0,68.
2. Pada penerapan skenario 1 tidak terdapat waktu tundaan karena tidak terjadi konflik
pada simpang. Sementara itu, bila diterapkan skenario 2 waktu tundaan menurun
tetapi tidak sebaik skenario 1.
3. Berdasarkan hasil scoring, skenario 1 memperoleh hasil tertinggi baik pada ruas jalan
maupun simpang.
Saran
Beberapa hal yang perlu diperhatikan dalam menganalisis dan memecahkan masalah lalu
lintas dengan menerapkan manajemen lalu lintas sistem satu arah, antara lain:
1. Pemilihan rute yang akan diterapkan dalam skenario sistem satu arah (SSA) dibuat
sedemikian rupa sehingga meminimalisir titik konflik yang terjadi.
2. Penerapan sistem satu arah (SSA) skenario 1 pada ruas Jl. Sultan Agung -
Jl.Sisingamangaraja - Jl. Dr. Wahidin dapat direkomendasikan. Hal ini dikarenakan
tidak terjadi konflik pada simpang sehingga waktu tundaan pada simpang berkurang
bahkan tidak ada. Selain itu, nilai DS ruas jalan pada skenario 1 dibandingkan dengan
kondisi eksisting dan skenario 2 lebih rendah dan penyebaran beban lalu lintas lebih
merata.
3. Perlu adanya analisis lebih lanjut, seperti analisis kelayakan, dampak ekonomi dan
sosial, serta biaya operasi kendaraan jika ruas tersebut diberlakukan sistem satu arah;
4. Perlu dalam studi lanjut adanya skenario dengan memprediksi perkembangan lalu
lintas untuk mengetahui sampai kapan sistem manajemen lalu lintas satu arah dapat
diberlakukan.
5. Perlu dilakukan analisis kembali setelah sistem ini diberlakukan yang berguna
sebagai pengontrol apakah sistem ini dapat berjalan dengan baik atau tidak.
6. Pemberian marka-marka dan rambu-rambu lalu lintas yang jelas agar pemakai jalan
mengerti dan patuh.
DAFTAR PUSTAKA
Badan Pusat Statistik. 2010. Hasil Sensus Penduduk 2010 Provinsi Jawa Tengah.
(http://www.bps.go.id/hasilSP2010/jateng/3300.pdf).
Badan Standarisasi Nasional. 2004. Standar Geometri Jalan Perkotaan, RSNI T-14-2004.
Direktorat Jenderal Bina Marga ± Direktorat Pembinaan Jalan Kota ± Departemen
Pekerjaan Umum, 1992. Standar Perencanaan Geometrik untuk Jalan Perkotaan.
Departemen Pekerjaan Umum Dirjen Bina Marga. 1997. Manual Kapasitas Jalan
Indonesia (MKJI).
Direktorat Jenderal Perhubungan Darat. Undang ± Undang Republik Indonesia Nomor 22
tahun 2009 Tentang Lalu Lintas dan Angkutan Jalan beserta Peraturan
Pelaksanaannya. Departemen Perhubungan, Jakarta.
152
Negara Republik Indonesia. 2004. Undang-Undang No. 38 tahun 2004 tentang Jalan.
Jakarta.
Negara Republik Indonesia. 2006. Undang-Undang No. 34 tahun 2006 tentang Jalan.
Jakarta.
Hoobs, F.D. 1995. Perencanaan dan Teknik Lalu Lintas. Yogyakarta : Gadjah Mada
University Press.
Wells, G.R. 1993. Rekayasa Lalu Lintas. Jakarta : Bharatara.
D.N., Agung., dan Fatkhurrahmat, I. 2011. Manajemen Lalu Lintas Satu Arah Kawasan
Utara dan Timur Kota Semarang. Tugas Akhir. Teknik Sipil Universitas
Diponegoro : Semarang
Junalia, A.N. dan Mahasin, A. 2013. Manajemen Lalu Lintas Jl.Tentara Pelajar ±
Jl.Kedungmundu Kota Semarang (Perempatan Cinde-Kantor PDAM
Kedungmundu). Tugas Akhir. Teknik Sipil Universitas Diponegoro : Semarang
Sukirman, S. 2000. Dasar-Dasar Perencanaan Geometrik Jalan. Bandung : Nova
----------. 1997. Sistem Transportasi. Jakarta : Gunadarma
Tamin, O.Z. 2000. Perencanaan dan Pemodelan Transportasi. Institut Teknologi
Bandung : Bandung.
Darmawan, A. dan Permana, A.A. 2013. Manajemen Lalu Lintas Satu Arah Kawasan
Timur Kota Semarang. Tugas Akhir. Teknik Sipil Universitas Diponegoro :
Semarang
Maulwy, O. dan Rendy, Putra. 2013. Manajemen Lalu Lintas Sistem Satu Arah pada Jalan
Pandanaran, Jalan MH.Thamrin, Jalan Gajah Mada, Jalan Depok, dan Jalan
Pemuda Kota Semarang. Tugas Akhir. Teknik Sipil Universitas Diponegoro :
Semarang
http://id.wikibooks.org, Manajemen Lalu Lintas, Diunduh pada tanggal 19 Mei 2013,
pukul 17.00 WIB
http://maps.google.co.id, Diunduh pada tanggal 19 Mei 2013, pukul 16.00 WIB
153
ANALISIS LALU LINTAS PENERAPAN SISTEM SATU ARAH
DI KAWASAN DUKUH ATAS, JAKARTA
Budi Hartanto Susilo1, Ivan Imanuel2

1
Guru Besar, Program Studi Teknik Sipil, Fakultas Teknik, Universitas Kristen Maranatha
2
Alumni, Program Studi Teknik Sipil, Fakultas Teknik, Universitas Kristen Maranatha
Email: budiharsus@yahoo.com, ivan.imanuel@yahoo.com
ABSTRAK
Dukuh Atas merupakan tempat bertemunya lima moda transportasi Jakarta di pusat kota. Dengan
masalah kemacetan yang parah pada jam sibuk, Dinas Perhubungan DKI Jakarta berencana
menerapkan manajemen dan rekayasa lalu lintas Sistem Satu Arah pada kawasan tersebut. Studi
ini membahas dampak lalu lintas penerapan SSA di Dukuh Atas dengan analisis mikrosimulasi
menggunakan Vissim. Hasil studi menunjukkan dengan adanya SSA, rata-rata kecepatan pada ruas
jalan meningkat dari 15,5 km/jam menjadi 17,7 km/jam dan tundaan pada simpang berkurang dari
60,3 detik/kendaraan menjadi 43,7 detik/kendaraan.
Kata Kunci: lalu lintas, manajemen dan rekayasa, sistem satu arah, Dukuh Atas, Vissim.
ABSTRACT
Dukuh Atas is an area where five modes of transportation in Jakarta meet in the city center. Due
to heavy traffic congestion problems on rush hours, the Department of Transportation of Jakarta
plans to implement traffic management called hone way system at the area. This study tells about
the impact of the one-way system on traffic at Dukuh Atas with microsimulation analysis using
Vissim. The result shows an increment of average speed on road sections from 15.5 km/hour to
17.7 km/hour and reduced average delays on road intersections from 60.3 seconds/vehicle to 43.7
seconds/vehicle.
Keywords: traffic engineering, traffic management, one-way system, Dukuh Atas, Vissim.
1. PENDAHULUAN
Kawasan Dukuh Atas merupakan tempat bertemunya MRT, LRT, KRL, kereta
bandara dan busway seperti terlihat pada
Gambar 1. Sebagai kawasan penting di pusat kota, Dukuh Atas mengalami kendala
kemacetan lalu lintas yang parah terutama pada jam sibuk. Kemacetan memang
merupakan masalah umum di Jakarta sejak beberapa tahun terakhir. Hal ini disebabkan
oleh transportasi yang berorientasi pada jalan dan kendaraan pribadi (Putranto, 2010).
Kerugian akibat kemacetan di sepanjang Jl. Jend. Sudirman, Jakarta diperkirakan
mencapai Rp 19 trilyun/tahun (Syaukat et al., 2015). Dalam rangka mengurai kemacetan
tersebut Dinas Perhubungan Provinsi DKI Jakarta berencana menerapkan manajemen dan
rekayasa lalu lintas sistem satu arah di kawasan Dukuh Atas.
Sistem Satu Arah (SSA) merupakan salah satu manajemen lalu lintas dengan cara
membuat jalan satu arah pada beberapa ruas jalan yang saling berhubung hingga
Analisis Lalu Lintas Penerapan Sistem Satu Arah Di Kawasan Dukuh Atas, Jakarta 105
(Budi Hartanto Susilo, Ivan Imanuel)
mengelilingi suatu wilayah. Dengan adanya SSA, diharapkan konflik kendaraan di
simpang-simpang berkurang sehingga pergerakan arus lalu lintas menjadi lebih lancar.
SSA sendiri sudah banyak diterapkan pada kota-kota besar di Indonesia, beberapa
contohnya di Tanah Abang, Sawah Besar, Kebayoran Baru, dan Jatinegara di Jakarta,
Pasirkaliki, Babakan Ciamis, dan Balonggede di Bandung, Petisah Tengah di Medan,
Krembangan dan Genteng di Surabaya, Karampuang di Makassar dan tempat-tempat
lainnya.
Sumber: PT MRT Jakarta (2018)

Gambar 1. Dukuh Atas, tempat bertemunya 5 moda transportasi Jakarta
Apakah SSA mampu meningkatkan kecepatan lalu lintas di kawasan Dukuh

Atas? Studi ini bertujuan untuk memperkirakan dampak lalu lintas penerapan SSA pada
kawasan Dukuh Atas menggunakan mikrosimulasi dengan perangkat lunak Vissim 11.
Studi ini merupakan penelitian eksperimen terhadap pergerakan lalu lintas menggunakan
model mikrosimulasi terhadap kondisi eksisting (do nothing) dan kondisi rencana (do
something). Beberapa studi serupa berkaitan analisis lalu lintas dengan mikrosimulasi
Vissim telah dipublikasikan. Irawan dan Putri (2015) menyusun kalibrasi model Vissim
pada simpang bersinyal dengan studi kasus Simpang Tugu, Yogyakarta. Negoro et al.
106 Jurnal Teknik Sipil Volume 14 Nomor 2, Oktober 2018 : 105-200

(2018) menganalisis pengaruh lima skenario manajemen kecepatan terhadap panjang
antrean pada gerbang tol Palimanan. Menggunakan Vissim juga, Wikayanti et al. (2018)
mengusulkan solusi peningkatan kinerja simpang dengan cara pengaturan siklus dan
pelebaran lengan.
2. METODOLOGI PENELITIAN
Studi analisis dampak lalu lintas penerapan kawasan Dukuh Atas ini secara garis
besar dimulai dari pengumpulan data, pemodelan lalu lintas, kalibrasi dan validasi,
analisis perbandingan, dan penyusunan kesimpulan dan rekomendasi. Tahapan kegiatan
studi ini dijelaskan dengan diagram alir pada Gambar 2.
Latar Belakang
Rumusan Masalah
Tujuan Penelitian
Pengumpulan Data Rencana Sistem

Satu Arah
Kecepatan Volume Inventarisasi
Pengolahan Data & Pemodelan
Model Eksisting Model Rencana

Jalan 2 Arah Jalan 1 Arah
Kinerja Lalu Lintas Eksisting: Kinerja Lalu Lintas Rencana:

Kecepatan Kecepatan
Tundaan Tundaan
Peningkatan Kinerja Tidak
Ya
Kesimpulan & Rekomendasi
Gambar 2. Diagram alir tahapan studi
Pemodelan lalu lintas dalam studi ini menggunakan pendekatan pembebanan lalu
lintas statis. Artinya pola pergerakan asal, tujuan, dan rute perjalanan lalu lintas dalam
kawasan yang ditinjau telah ditentukan sebelumnya dan tidak berubah selama simulasi.
Analisis Lalu Lintas Penerapan Sistem Satu Arah Di Kawasan Dukuh Atas, Jakarta107
Kalibrasi perilaku lalu lintas dan validasi model lalu lintas berdasarkan parameter volume
dan kecepatan perjalanan. Validasi model menggunakan metode R2. Pembatasan wilayah
studi juga perlu dilakukan agar analisis dapat lebih terfokus pada dampak lalu lintas
penerapan SSA. Jaringan jalan yang ditinjau antara lain Jl. Jenderal Sudirman, Jl. R.M
Margono Djojohadikusumo, Jl. Galunggung, Jl. Setia Budi Tengah, Ramp Barat, Ramp
Timur, sisi barat Landmark dan sisi timur Landmark.
Model rencana terdiri atas pengaturan lalu lintas seperti terlihat pada Gambar 3,
dengan rincian:
1. sisi selatan (Jl. Setia Budi Tengah) menjadi satu arah dari timur ke barat, atau dari
arah Manggarai menuju Dukuh atas;
2. sisi barat Landmark menjadi satu arah dari selatan ke utara, atau menuju Jalan
Galunggung maupun Jl. R.M Margono Djojohadikusumo;
3. sisi utara (Jl. Galunggung) satu arah dari barat ke timur, atau dari arah Dukuh Atas
menuju Manggarai;
4. sisi timur Landmark menjadi satu arah dari utara ke selatan.
Eksisting Rencana
Gambar 3. Kondisi Eksisting dan Rencana SSA di Dukuh Atas
1. PENGUMPULAN DATA
Untuk membuat model lalu lintas, survei inventarisasi jalan terlebih
dahulu dilakukan. Survei inventarisasi jalan mengikuti Tata Cara No.
017/T/BNKT/1990. Survei ini meliputi pendataan ruas jalan dan simpang, tipe
jalan, lebar jalan, pengaturan lalu lintas (rambu, marka, median, separator, dan
lain-lain). Tabel 1 menunjukkan hasil inventarisasi jalan pada 14 ruas jalan di
kawasan Dukuh Atas, Jakarta.
Pengumpulan data volume lalu lintas sesuai dengan Pd. T-19-2004-B
tentang survei pencacahan lalu lintas dengan cara manual, pada hari kerja jam

sibuk sore hari saat lalu lintas mengalami puncak kepadatan, dengan hasil
sebagaimana terlihat pada Tabel 2 Pada saat yang bersamaan juga dilakukan
pengukuran waktu perjalanan lalu lintas untuk menyatakan kinerja lalu lintas
secara aktual. Survei waktu perjalanan mengikuti Panduan No. 001/T/BNKT/1990
menggunakan metode kendaraan contoh (floating car method).
Tabel 1. Inventarisasi Ruas Jalan di Kawasan Dukuh Atas

Lebar
Tipe Hambatan
No. Nama Jalan Lajur Bahu Trotoar
Jalan Samping
(m) (m) (m)
1 Jl. Jenderal Sudirman 1
a. Arah Bundaran HI (Utara) 3,46 0,5 9,5 Tinggi
10/2D
b. Arah Senayan (Selatan) 3,46 0,5 9,5 Tinggi
2 Jl. Jenderal Sudirman 2 (jembatan)
a. Arah Bundaran HI (Utara) 3,1 0,5 5 - 9,5 Tinggi
10/2D
b. Arah Senayan (Selatan) 3,4 0,5 5 - 9,5 Tinggi
3 Jl. RM Margono Djojohadikusumo
a. Arah Tanah Abang (Barat) 3,3 0,5 1,5-3 Rendah
4/2D
b. Arah Manggarai (Timur) 3 0,5 1,5-3 Sedang
4 Jl. Galunggung 1 (underpass)
a. Arah Tanah Abang (Barat) 3,5 0,5 1,5-3 Rendah
4/2D
b. Arah Manggarai (Timur) 3,5 0,5 1,5-3 Sedang
5 Jl. Galunggung 2 (utara Landmark)
a. Arah Tanah Abang (Barat) 3 0,5 0,5 Rendah
7/2D
b. Arah Manggarai (Timur) 3 0,5 1,5 Sedang
6 Jl. Galunggung 3 (utara waduk Setia Budi)
a. Arah Tanah Abang (Barat) 3 0,5 1,5 Sedang
6/2D
b. Arah Manggarai (Timur) 3 0,5 0,5 Rendah
7 Jl. Setia Budi Tengah 1 (selatan Landmark)
a. Arah Sudirman (Barat) 2,8 - 1 Tinggi
5/2D
b. Arah Manggarai (Timur) 2,6 - 1 Sedang
8 Jl. Setia Budi Tengah 2 (selatan waduk Setia Budi)
a. Arah Sudirman (Barat) 3,5 2 Rendah
4/2D
b. Arah Manggarai (Timur) 3,5 2 Rendah
9 Jl. Sisi Barat Landmark
a. Arah underpass (Utara) 3,25 - 1 Sedang
4/2D
b. Arah Sudirman (Selatan) 3,5 0,5 1,5 Tinggi
10 Jl. Sisi Timur Landmark
a. Arah Galunggung (Utara) 3,5 - 1,5 Rendah
4/2D
b. Arah Setia Budi (Selatan) 3,5 1 2 Sedang
11 Ramp Sudirman Barat
a. Arah Margono (keluar) 2/1 3,5 - 1,5-3 Sedang
b. Arah Sudirman (masuk) 2/1 3,5 - 1,5-3 Sedang
12 Ramp Sudirman Timur
a. Arah Setia Budi (keluar) 2/1 4,5 - 0,5 Sedang
b. Arah Sudirman (masuk) 2/1 3,75 - 1,5 Sedang
13 Jl. Karet Pasar Baru Timur III
a. Arah Mas Mansyur (ke Barat) 4,5 - 0,5 Sedang
2/2UD
b. Arah Sudirman (ke Timur) 4,5 - 0,5 Tinggi
Pengendalian lalu lintas di simpang dengan APILL merupakan salah satu
input pada model sehingga memerlukan pengukuran waktu sinyal untuk setiap
pendekat. Terdapat 2 simpang ber-APILL di Dukuh Atas, yaitu simpang Ramp
Sudirman Barat-Margono Djojohadikusumo (Dukuh Bawah) dan simpang
Galunggung-Sisi Timur Landmark (Galunggung).
Tabel 2. Volume Lalu Lintas di Kawasan Dukuh Atas

Volume
Ruas Jalan Arah SM KR KB KTB
(kend/jam)
Sudirman U-S 6767 4719 1920 121 7
Sudirman S-U 6330 3732 2479 116 3
Thamrin U-S 8249 4662 3534 45 8
Thamrin S-U 4659 3034 1502 118 5
Galunggung B-T 4063 3394 616 42 11
Galunggung T-B 6396 5466 862 56 12
Setia Budi U-S 2890 1788 1094 8 0
Setia Budi S-U 2167 1455 709 3 0
Margono B-T 4756 3726 1003 19 8
Margono T-B 4517 3179 1309 24 5
Karet Pasar Baru III B-T 1331 859 465 7 0
Karet Pasar Baru III T-B 509 326 181 2 0
SM = Sepeda Motor, KR = Kendaraan Ringan, KB = Kendaraan Berat, KTB = Kendaraan Tak Bermotor
4. PEMODELAN LALU LINTAS DAN ANALISIS

Hasil output model eksisting dari Vissim divalidasi dengan data survei
lapangan, seperti terlihat pada Gambar 4. Nilai R2 yang mencapai 0,893
menunjukkan konsistensi model dengan hasil lapangan dan dengan demikian
model dapat digunakan. Selanjutnya, model lalu lintas tersebut dimodifikasi
dengan mengubah arah lalu lintas di sekeliling Landmark menjadi satu arah
berlawanan arah jarum jam.
Perbandingan hasil kinerja lalu lintas pada kondisi eksisting dengan
rencana SSA ditunjukkan pada Tabel 3 dan Tabel 4. Rata-rata kecepatan lalu
lintas di ruas jalan adalah 15,6 km/jam dengan tingkat pelayanan (LOS) E.
Kinerja lalu lintas pada simpang berkisar antara D dan F, dengan tundaan rata-rata
sebesar 60,3 detik/kendaraan.

Gambar 4. Validasi Model Vissim
Gambar 5. Model Simulasi Lalu Lintas Eksisting Kawasan Dukuh Atas
Tabel 3. Perbandingan Kinerja Ruas Jalan Eksisting dengan Rencana SSA

Kondisi Eksisting Rencana SSA
Volume
No Ruas Kec Kec
(kend/jam) LOS LOS
(km/jam) (km/jam)
1 Sudirman - SU 6767 22,7 E 22,7 E
2 Sudirman - US 6330 7,1 F 7,2 F
3 Thamrin - SU 4659 21,8 E 18,8 E
4 Thamrin - US 8249 10,6 E 12,8 E
5 Margono - BT 4756 20,8 E 43,2 E
Kondisi Eksisting Rencana SSA
Volume
No Ruas Kec Kec
(kend/jam) LOS LOS
(km/jam) (km/jam)
6 Margono - TB 4517 15,1 E 16,5 E
7 Galunggung - BT 4063 22,7 E 22,1 E
8 Galunggung - TB 6396 10,3 E 7,4 F
Rata-rata 15,6 E 17,7 E
Tabel 4. Perbandingan Kinerja Simpang Jalan Eksisting dengan Rencana SSA

Eksisting SSA Timur
No Simpang Tundaan Tundaan
LOS LOS
(detik/kend) (detik/kend)
1 Dukuh Bawah 35,7 D 15,56 B
2 Galunggung 89,4 F 61,92 F
3 Landmark Barat 55,9 F 53,6 F
Rata-rata 60,3 F 43,7 E
Hasil pembebanan lalu lintas dengan rencana SSA, seperti terlihat pada
menunjukkan terjadinya peningkatan signifikan di Jl. Margono arah ke timur dari
20,8 km/jam menjadi 43,2 km/jam, sementara pada ruas jalan lainnya peningkatan
yang terjadi kurang dari 2 km/jam dan penurunan kecepatan terjadi di Jl.
Galunggung arah ke barat dan Jl. Thamrin arah ke utara.
Secara keseluruhan kinerja ruas jalan meningkat dengan rata-rata
kecepatan lalu lintas menjadi 17,7 km/jam. Dampak penerapan SSA terhadap
kinerja simpang adalah positif pada seluruh simpang. Tundaan pada ketiga
simpang yang ditunjau menurun secara signifikan dengan tundaan rata-rata
menjadi 43,7 detik/kendaraan.

Gambar 6. Model Lalu Lintas Sistem Satu Arah di Kawasan Dukuh Atas
5 KESIMPULAN DAN REKOMENDASI

Kinerja lalu lintas kondisi eksisting di Kawasan Dukuh Atas memiliki
kecepatan rata-rata pada ruas jalan sebesar 15,6 km/jam dan tundaan rata-rata
pada simpang sebesar 60,3 detik/kendaraan. Dengan penerapan SSA, kecepatan
rata-rata meningkat menjadi 17,7 km/jam dan tundaan di simpang turun menjadi
43,7 detik/kendaraan.
Penerapan SSA diperkirakan akan meningkatkan kinerja lalu lintas di
kawasan Dukuh Atas. Penerapan SSA ini akan lebih baik jika diterapkan juga
pada kawasan Dukuh Atas bagian barat mengelilingi KCP BNI sehingga simpang
Dukuh Bawah tidak perlu menggunakan APILL.
DAFTAR PUSTAKA
1. Departemen Permukiman dan Prasarana Wilayah. (2004). Survai Pencacahan

Lalu Lintas dengan cara Manual. Jakarta: Departemen Permukiman dan
Prasarana Wilayah.
2. Direktorat Pembinaan Jalan Kota. (1990). Panduan Survai dan Perhitungan
Waktu Perjalanan Lalu Lintas. Jakarta: Direktorat Jenderal Bina Marga.
3. Irawan, M. Z., & Putri, N. H. (2015). Kalibrasi Vissim Untuk Mikrosimulasi
Arus Lalu Lintas Tercampur Pada Simpang Bersinyal (Studi Kasus: Simpang
Tugu, Yogyakarta). Jurnal Penelitian Transportasi Multimoda, 13(03), 97-106.
4. Negoro, Y. A., Munawar, A., & Irawan, M. Z. (2018). Analisis Pengaruh
Manajemen Kecepatan Terhadap Antrian Kendaraan Pada Exit Gerbang Tol
Periode Liburan. Jurnal Penelitian Transportasi Darat, 20(1), 33-48.
doi:10.25104/jptd.v20i1.649
5. PT MRT Jakarta. (2018). Laporan Final Usulan Panduan Rancang Kota
Kawasan TOD Dukuh Atas. Jakarta: PT MRT Jakarta.
6. Putranto, L. S. (2010). The Evaluation of Space Mean Speeds of Road Links
Surrounding New Developments in Jakarta. The Seventh Asia Pacific Conference
on Transportation and the Environment Proceedings.
7. Syaukat, Y., Sarma, M., Falatehan, A. F., & Bahtiar, R. (2015). Analysis of
Willingness to Pay (WTP) to Determine Road Pricing in Jakarta. Scientific
Journal of PPI-UKM, 2(6), 258-260.
8. Wikayanti, N., Azwansyah, H., & Kadarini, S. N. (2018). Penggunaan Software
Vissim Untuk Analisis Simpang Bersinyal (Studi Kasus Jalan Sultan Hamid II –
Jalan Gusti Situt Mahmud –Jalan 28 Oktober – Jalan Selat Panjang). Jurnal
Mahasiswa Teknik Sipil Universitas Tanjungpura, 5(3), 1-11. Retrieved from
http://jurnal.untan.ac.id/index.php/JMHMS/article/view/30477/75676579617..

Jurnal Teknik Sipil Vol. IV, No. 2, September 2015
KAJIAN PENERAPAN REKAYASA LALU LINTAS SISTEM SATU

ARAH PADA SIMPANG TIGA STRAAT A KOTA KUPANG
Margareth E. Bolla1 (mgi_ub 08@yahoo.com)
Yunita A. Messah2 (yunitamessah@gmail.com)
Lauren Johanes3 (laurenjohanes08@gmail.com)
ABSTRAK
Hasil penelitian simpang tiga bersinyal Straat A Kota Kupang yang dilakukan oleh Frans (2014),
menyimpulkan bahwa kondisi kinerja simpang berada pada tingkat pelayanan F. Untuk
mengatasi masalah tersebut maka pada penelitian ini dilakukan rekayasa lalu lintas berupa
penerapan sistem lalu lintas satu arah pada simpul simpang tiga Straat A tersebut, yang meliputi
ruas jalan A.Yani, Jalan Flores, dan Jalan Sumba. Pengolahan dan analisa data menggunakan
MKJI 1997. Hasil penelitian menunjukkan kinerja Simpang Tiga Straat A dengan penerapan
sistem lalu lintas satu arah dikategorikan dalam tingkat pelayanan B yaitu arus stabil, kepadatan
rendah, pengemudi masih punya cukup kebebasan memilih kecepatan. Kinerja lalulintas pada
ruas jalan yang dipengaruhi adalah ruas jalan A. Yani dengan volume kendaraan maksimum jam
puncak 1.646 smp/jam, hambatan samping tergolong sedang, kecepatan aktual 48 km/jam,
derajat kejenuhan 0,48 dan tingkat pelayanan C; Pada ruas jalan Sumba, volume kendaraan
maksimum jam puncak 1.509 smp/jam, hambatan samping sedang, kecepatan aktual 50 km/jam,
derajat kejenuhan 0,48 dan tingkat pelayanan C; Ruas Jalan Flores, volume kendaraan
maksimum jam puncaknya adalah 1.342,9 smp/jam, hambatan samping sedang, kecepatan aktual
44 km/jam, derajat kejenuhan 0,48 serta tingkat pelayanan C. Sehingga dapat disimpulkan bahwa
penerapan sistem lalu lintas satu arah pada simpang tiga Straat A layak dilakukan.
Kata Kunci: Simpang, Tingkat Pelayanan, Sistem Lalu Lintas Satu Arah
ABSTRACT
Based on the previous research on the three-legged signalized intersections on Straat A in

Kupang City (Frans, 2014), the result showed that the performance was at F level of service. To
overcome this problem, this research was conducted by apply the traffic engineering that is the
one-way traffic system on the Straat A intersection that covers A. Yani, Flores and Sumba Roads.
Data processing and analysis were done using the Indonesian Highway Capacity Manual
(MKJI, 1997). The result showed that the performance of Straat A intersection with the
implementation of one-way traffic system is categorized at Level of Service (LoS) B that is has
steady traffic flow, low traffic density, and the driver still has enough freedom to choose the
speed. The traffic performance of the roads that were affected by one way traffic system are as
follow: 1. On A. Yani road, the maximum peak hour volume is 1.646 pcu/hour, the side friction is
classified medium, the actual speed is 48 kph, the degree of saturation is 0.48 and LoS is at C
level; 2. On Sumba road, the maximum peak hour volume is 1509 pcu/hour, the side friction is
classified medium, the actual speed is 50 kph, the degree of saturation is 0.48, and LoS level is at
grade C; 3. On Flores Road, the maximum peak hour volume is 1.342,9 pcu/hour, the side
friction is also classified medium, the actual speed is 44 kph, the degree of saturation is 0.48 and
the LoS level is at grade C too. Thus, it can be concluded that the implementation of one-way
traffic system on the three-legged intersection Straat A is feasible
Keywords: Intersection, Level of Service, One-Way Traffic System
1
Dosen pada Jurusan Teknik Sipil, FST Undana;
2
Dosen pada Jurusan Teknik Sipil, FST Undana;
3
Penamat dari Jurusan Teknik Sipil, FST Udana
Bolla, M. E., et.al., “Kajian Penerapan Rekayasa Lalu Lintas Sistem Satu Arah pada Simpang Tiga Straat A Kota Kupang”
217
PENDAHULUAN
Kota Kupang sebagai ibu kota propinsi Nusa Tenggara Timur (NTT) merupakan kota yang
sedang berkembang, hal ini ditandai dengan adanya pembangunan di berbagai tempat di sekitar
Kota Kupang. Perkembangan Kota Kupang ditandai pula dengan meningkatnya pengguna
kendaraan bermotor, di mana jumlah kendaraan pribadi (sepeda motor dan mobil) lebih banyak
jika dibandingkan dengan jumlah angkutan umum sehingga menyebabkan peningkatan volume
lalu lintas (Dishub Kota Kupang, 2014).
Berdasarkan hasil dari penelitian sebelumnya (Frans, 2014). Besarnya volume lalu lintas pada
jam-jam sibuk menyebabkan tingkat pelayanan simpang tiga bersinyal Straat A berada pada
level F (kondisi terburuk). Hal ini disebabkan karena adanya perubahan intensitas tata guna
lahan di sekitar simpang dan dominannya penggunaan kendaraan pribadi, antrian dan tundaan
yang sangat besar. Untuk mengatasi masalah tersebut, maka berdasarkan penelitian tersebut
disarankan untuk dilakukan rekayasa lalu lintas berupa penerapan sistem satu arah pada simpang
tiga bersinyal Straat A.
Adapun tujuan dari penelitian ini adalah untuk mengetahui kinerja simpang tiga Straat A dengan
diterapkannya simulasi rekayasa lalu lintas satu arah serta untuk mengetahui kinerja ruas-ruas
jalan pada simpul simpang tiga Straat A akibat diterapkannya simulasi rekayasa lalulintas satu
arah.
TINJAUAN PUSTAKA
Simpang
Simpang adalah bagian dari sistem jaringan jalan yang mempertemukan dua ruas jalan atau lebih
dari dua ruas jalan, secara umum kapasitas simpang dapat dikontrol dengan mengendalikan
volume lalu lintas dan sistem jaringan tersebut.
Alat pengendali lalu lintas meliputi: rambu, marka, penghalang yang dapat dipindahkan, dan
lampu lalu lintas. Seluruh peralatan pengendali lalu lintas pada simpang dapat digunakan secara
terpisah atau digabungkan bila perlu (Khisty dan Kent, 2005).
Sistem Satu Arah

Sistem satu arah adalah suatu pola lalu lintas yang dilakukan dengan merubah jalan dua arah
menjadi jalan satu arah yang berfungsi untuk meningkatkan keselamatan dan kapasitas jalan dan
persimpangan sehingga meningkatkan kelancaran lalu lintas yang biasanya diterapkan di
wilayah perkotaan, Manfaat dari penerapan jalan satu arah adalah untuk meningkatkan kapasitas
dan meningkatkan keselamatan (Wikipedia).
Karakteristik Jalan Satu Arah

Tipe jalan ini meliputi semua jalan satu arah ketentukan sebagai berikut:
a. Lebar jalur lalu lintas 5 – 10,5 m.
b. Lebar bahu efektif paling sedikit 2 m pada setiap sisi.
c. Tidak ada median.
d. Hambatan samping rendah.
e. Tipe alinyemen datar. (Wikipedia)
Volume Lalu Lintas

Volume adalah jumlah kendaraan yang melintasi suatu ruas jalan pada periode waktu tertentu,
diukur dalam satuan kendaraan per satuan waktu. Volume lalu lintas pada suatu jalan akan
bervariasi tergantung pada volume total dua arah, arah lalu lintas, volume harian, bulanan dan
tahunan pada komposisi kendaraan. (Khisty dan Kent, 2005).
Kinerja Lalu Lintas
MKJI 1997 menyatakan ukuran kinerja lalu lintas diantaranya adalah tingkat pelayanan
merupakan parameter yang menerangkan kondisi operasional dari fasilitas lalu lintas seperti
218
yang dinilai oleh pembina jalan yang bisa dinyatakan dalam kapasitas (C) atau derajat kejenuhan
(DS) atau kecepatan rata-rata (V), atau waktu tempuh (TT), atau tundaan (D), atau peluang
antrian (QP), atau panjang antrian atau rasio kendaraan terhenti atau pula kombinasi dari syarat
yang ada.
Untuk kinerja simpang tak bersinyal parameter yang mempengaruhi geometri, lingkungan
dan lalu lintas yaitu Kapasitas (C), Derajat kejenuhan (DS), Tundaan (D), Peluang antrian (QP).
Kapasitas Simpang Tak Bersinyal

MKJI (1997) mendefinisikan bahwa kapasitas adalah arus lalu lintas maksimum yang dapat
dipertahankan (tetap) pada suatu bagian jalan dalam kondisi tertentu dinyatakan dalam
kendaraan/jam atau smp/jam.
Kapasitas total suatu persimpangan dapat dinyatakan sebagai hasil perkalian antara kapasitas
dasar (Co) dan faktor-faktor penyesuaian (F). Rumusan kapasitas simpang menurut MKJI 1997
dituliskan sebagai berikut:
C = Co × FW × FM × FCS × FRSU × FLT×FRT × FMi (1)
Dimana : C : Kapasitas (Smp/jam)
C0 : Kapasitas dasar
FW : Faktor Penyesuaian lebar masuk
FM : Faktor penyesuaian tipe median jalan utama
FCS : Faktor penyesuaian ukuran kota
FRSU :Faktor penyasuaian tipe lingkungan jalan, hambatan samping dan
kendaraan tak bermotor
FLT : Faktor penyesuaian belok kiri
FRT : Faktor penyesuaian belok kanan
FMi : Faktor penyesuaian ruas arus jalan minor
Lebar Pendekat dan Tipe Simpang

Lebar pendekat diukur pada jarak 10 m dari garis imajiner yang menghubungkan tepi perkerasan
dari jalan berpotongan. Tipe simpang menentukan jumlah lengan simpang dan jumlah lajur pada
jalan utama dan jalan minor pada simpang tersebut dengan kode tiga angka
Kapasitas Dasar
Kapasitas dasar adalah kapasitas segmen jalan untuk suatu kondisi yang ditentukan sebelumnya
(geometri, pola arus lalu lintas, dan faktor lingkungan).
Tabel 1 Nilai Kapasitas Dasar (CO)
Tipe Simpang IT Kapasitas Dasar
smp/jam
322 2700
342 2900
324 dan 344 3200
333 3300
422 2900
424 dan 444 3400
Sumber: MKJI, 1997
Faktor Penyesuaian Median Jalan Utama

Median disebut lebar jika kendaraan ringan standar dapat berlindung pada daerah median tanpa
mengganggu arus berangkat pada jalan utama. Hal ini mungkin terjadi jika lebar median 3 m
atau lebih. Pada beberapa keadaan, misalnya jika pendekat jalan utama lebar, hal ini mungkin
terjadi jika median lebih sempit.
219
Faktor Penyesuaian Ukuran Kota

Tabel 3 Faktor penyesuaian ukuran kota (FCS)
Ukuran Kota CS Penduduk Juta Faktor Penyesuaian Ukuran Kota
FCS
Sangat Kecil <1 0.82
Kecil 0.1 – 0.5 0.88
Sedang 0.5 – 1.0 0.94
Besar 1.0 – 3.0 1
Sangat Besar >3 1.05
Sumber: MKJI, 1997
Faktor Penyesuaian Tipe Lingkungan Jalan, Hambatan Samping dan Kendaraan Tak
Bermotor
Faktor penyesuaian tipe lingkungan jalan, hambatan samping dan kendaraan tak bermotor, FRSU
dihitung dengan menggunakan variabel masukan adalah tipe lingkungan jalan RE, kelas
hambatan samping SF dan rasio kendaraan tak bermotor UM/MV (PUM).
Tabel 4 Faktor Penyesuaian Tipe Lingkungan Jalan, Hambatan Samping dan Kendaraan Tak
Bermotor (FRSU)
Kelas tipe Kelas Hambatan Rasio kendaraan tak bermotor
lingkungan Samping SF PUM
Jalan RE 0,00 0,05 0,10 0,20 ≤ 0,25
Komersial Tinggi 0.93 0.88 0.84 0.74 0.70

Sedang 0.94 0.89 0.85 0.75 0.70
Rendah 0.95 0.90 0.86 0.76 0.71
Pemukiman Tinggi 0.96 0.91 0.86 0.77 0.72
Sedang 0.97 0.92 0.87 0.77 0.73
Rendah 0.98 0.93 0.88 0.78 0.74
Akses Tinggi/ sedang/ rendah 100 0.95 0.90 0.80 0.75
terbatas
Sumber: MKJI, 1997
Faktor Penyesuaian Belok Kiri

Faktor penyesuaian belok-kiri dirumuskan sebagai berikut:
Simpang 4-lengan FLT = 1.0 (2)
Simpang 3-lengan FLT = 0.84 + 1.61 PLT (3)
Faktor Penyesuaian Belok Kanan

Faktor penyesuaian belok-kanan sebagai berikut:
simpang 4-lengan FRT = 1,0 (4)
simpang 3-lengan FLT = 1,09 + 0,922 PRT (5)
Faktor Penyesuaian Rasio Arus Jalan Minor

Faktor penyesuaian rasio arus jalan minor ditentukan oleh variabel masukan adalah rasio arus
jalan minor (PMI) dan tipe simpang IT. Batas nilai yang diberikan untuk PMI pada gambar
adalah rentang dasar empiris dari manual.
Derajat Kejenuhan
Derajat kejenuhan (DS) merupakan rasio arus lalu lintas (smp/jam) terhadap kapasitas
(smp/jam), dapat ditulis dengan persamaan sebagai berikut:
DS = (6)
Dimana : DS : Derajat Kejenuhan
220
Q : Volume Kendaraan (Smp/jam)

C : Kapasitas (Smp/jam)
Tundaan (D)
Tundaan di persimpangan adalah total waktu hambatan rata-rata yang dialami oleh kendaraan
sewaktu melewati suatu simpang. Nilai tundaan mempengaruhi nilai waktu tempuh kendaraan.
Semakin tinggi nilai tundaan, semakin tinggi pula waktu tempuh.
Tundaan lalu lintas rata-rata untuk seluruh simpang (DTi) (detik/smp)

a. Untuk DS ≤ 1.0
DTi = 2 + (8.2078 × DS) – DS) × 2 (7)
b. Untuk DS > 1.0
.
DT = − 1 − DS × 2 (8)
. . ×
Dimana : DTi : Tundaan lalu lintas seluruh simpang

DS : Derajat Kejenuhan
Tundaan lalu lintas rata-rata untuk jalan major (DTMA)

a. Untuk DS ≤ 1.0
DTMA = 1.8 + (5.823 × DS) [(1-– DS) × 1.8 (9)
b. Untuk DS > 1.0
. #
DT!" = .# $ . × − 1 − DS × 1.8 (10)
Dimana : DTMA : Tundaan lalu lintas jalan mayor
Tundaan lalu lintas rata-rata jalan minor (DTMI)

Tundaan lalu lintas rata-rata jalan minor ditentukan berdasarkan tundaan lalu lintas rata-rata
(DTi) dan tundaan lalu lintas rata-rata jalan major (DTMA).
*+,-./ ×012 3 ,45 ×0145 6
&'() = ,47
(11)
Dimana : DTMI : Tundaan lalu lintas jalan minor
DTMA : Tundaan lalu lintas jalan mayor
Tundaan geometrik simpang (DG)

1. Untuk DS ≤ 1.0
DG= (1-DS) × (PT × 6 +(1-PT)× 3) + DS × 4 (12)
2. Untuk DS > 1.0
DG = 4 detik (13)
Dimana : DG : Tundaan geometri simpang
PT : Rasio belok total
Tundaan simpang (D)

D = DG + DTi (14)
Dimana : DG : Tundaan lalu lintas jalan minor
DTi : Tundaan lalu lintas jalan mayor
Peluang Antrian (QP%)

Batas atas:
QPa = (47.71 × DS) – (24.68 × DS2) + (56.47 × DS2) (15)
221
Batas bawah:
QPb = (9.02 × DS) + (20.66 × DS2) + (10.49 × DS2) (16)
Dimana : QPa : Peluang antrian batas atas

QPb : Peluang antrian batas bawah
Klasifikasi untuk ruas jalan

Klasifikasi jalan yang paling sederhana adalah membagi menjadi jalan umum (kecepatan/volume
tinggi) dan jalur minor (akses tinggi). Klasifikasi jalan menurut Peraturan Pemerintah No. 26
Tahun 1985 dikelompokkan menjadi :
1. Jalan arteri primer
2. Jalan kolektor primer
3. Jalan lokal primer
Hambatan Samping
Hambatan samping diperoleh dari pengamatan visual di lapangan. Kelas hambatan samping
dibedakan atas kategori rendah, sedang, tinggi
Kecepatan Arus bebas

Kecepatan arus bebas (FV) adalah kecepatan teoritis lalu lintas ketika kepadatan arus mendekati
nol. Untuk kecepatan arus bebas sesungguhnya dipakai persamaan berikut :
FV = (FVo + FVw) x FFVsf x FFVcs (17)
Dimana : FV : Kecepatan arus bebas kend. Ringan (Km/jam)
FV0 : Kecepatan arus bebas dasar kend. Ringan (Km/jam)
FVW : Penyesuaian lebar jalur lalu lintas (Km/jam)
FFVsf : Faktor penyesuaian hambatan samping
FFVcs : Faktor penyesuaian ukuran kota
Kapasitas Jalan
Kapasitas didefinisikan sebagai arus maksimum melalui suatu titik dijalan yang dapat
dipertahankan per satuan jam. Untuk jalan dengan banyak lajur, arus dipisahkan per arah dan
kapasitas ditentukan per lajur.
Persamaan dasar untuk menentukan kapasitas adalah sebahgai berikut :
C =C0 x FCW x FCSP x FCSF x FCCS (18)
Dimana : C : Kapasitas (Smp/jam)
C0 : Kapasitas dasar (Smp/jam)
FCW : Faktor Penyesuaian lebar jalur lalu lintas (Km/jam)
FCsp : Faktor penyesuaian pemisah arah
FCsf : Faktor penyesuaian hambatan samping
FCcs : Faktor penyesuaian ukuran kota
Tabel 5 Kapasitas Dasar Co Untuk Jalan Perkotaan

Kapasitas dasar
Tipe Jalan Catatan
(SMP/jam)
Empat lajur tebagi atau 1650 Per lajur
jalan satu arah
Empat lajur tak terbagi 1500 Per lajur
Dua lajur tak terbagi 2900 Total dua arah
Sumber: MKJI 1997
Untuk menentukan nilai FCw perlu diketahui terlebih dahulu nilai lebar efektif jalan tinjauan
yang kemudian dengan menggunakan tabel dibawah ini dapat kita dapatkan nilai FCw:
222
Tabel 6 Faktor Penyesuaian Kapasitas Untuk Pemisah Arah (FCsp)

Pemisah arah
50-50 60-40 70-30 80-20 90-10 100-0
SP %-%
Dua lajur
1,00 0,94 0,88 0,82 0,76 0,70
2/2
FCsp
Empat lajur
1,00 0,97 0,94 0,91 0,88 0,85
4/2
Sumber: MKJI 1997
Untuk jalan terbagi dan jalan satu-arah, faktor penyesuaian kapasitas untuk pemisah arah tidak
dapat diterapkan dan nilai 1,0. Kemudian penentuan nilai FCSF pengaruh hambatan samping dan
lebar bahu jalan berdasarkan tingkat hambatan samping dan lebar bahu jalan efektif.
Tabel 7 Faktor Penyesuaian FCcs Untuk Pengaruh Ukuran Kota
Pada Kapasitas Jalan Perkotaan
Ukuran Kota (Juta Penduduk) Faktor Penyesuaian untuk ukuran kotaFCcs
<0,1 0,86
0,1-0,5 0,90
0,5-1,0 0,94
1,0-3,0 1,00
>3,0 1,04
Sumber: MKJI 1997
Derajat Kejenuhan
Derajat kejenuhan (DS) didefinisikan sebagai rasio arus terhadap kapasitas digunakan sebagai
faktor utama dalam penentuan tingkat kinerja simpang dan segmen jalan.
Kecepatan
Kecepatan tempuh didefenisikan sebagai kecepatan rata-rata ruang dari kendaraan ringan (LV)
sepanjang segmen jalan:
V = L / TT (19)
Dimana : V : Kecepatan
L : Panjang segmen jalan
TT : Waktu tempuh
Waktu Tempuh (TT)

TT = L / V (20)
METODOLOGI PENELITIAN
Lokasi Penelitian
Penelitian ini dilakukan di Kota Kupang, Kecamatan Kota Lama, Kelurahan Oeba, yaitu pada 4
titik lokasi penelitian yakni: Titik 1: Simpang Straat A; Titik 2: Jl. Flores; Titik 3: Jl. A. Yani;
Titik 4: Jl. Sumba.
Waktu Penelitian
Hari Senin, 27 Oktober 2014 sampai Hari Sabtu, 01 November 2014 dari jam 07.00-14.00 WITA
dilanjutkan Jam 16.00-18.00 WITA.
223
Data Primer
Data primer berupa kondisi geometri, volume lalu lintas dan hambatan samping.
Data Sekunder
Data sekunder berupa penelitian terdahulu dan data jumlah penduduk.
HASIL DAN PEMBAHASAN

Analisis dan Pembahasan Simpang Straat A
Berdasarkan hasil survei yang telah dilakukan selama 6 hari, data-data tersebut kemudian
dianalisis. Adapun hasil analisis data survei tersebut adalah sebagai berikut:
1. Volume Lalu Lintas
Dari hasil rekapitulasi seperti terlihat pada gambar.1 diperoleh total volume kendaraan
terbesar: 2.803 smp/jam pada hari Sabtu, 01 November 2014 pukul 16.45–17.45 WITA.
Hasil survei juga menunjukan volume lalu lintas pada masing-masing lengan simpang.
Volume lalu lintas pada lengan simpang mayor D sebesar 1.378 smp/jam kemudian yang di
lengan simpang minor C sebesar 1.425 smp/jam dan volume lalu lintas pada lengan simpang
mayor B sebesar 1.660 smp/jam.
224
Volume Jam-Jaman Pada Hari Sabtu di Simpang STRAT A
2803
2798
2794
2793
2786
2779
2771
3000
2769
2767
2758
2752
2750
2746
2744
2739
2737
2728
2724
2708
2705
2703
2685
2669
2664
2614
2606
2582
2750
2501
2464
2424
2500
2250
2000
07.00-08.00
07.15-08.15
07.30-08.30
07.45-08.45
08.00-09.00
08.15-09.15
08.30-09.30
08.45-09.45
09.00-10.00
09.15-10.15
09.30-10.30
09.45-10.45
10.00-11.00
10.15-11.15
10.30-11.30
10.45-11.45
11.00-12.00
11.15-12.15
11.30-12.30
11.45-12.45
12.00-13.00
12.15-13.15
12.30-13.30
12.45-13.45
13.00-14.00
16.00-17.00
16.15-17.15
16.30-17.30
16.45-17.45
17.00-18.00
Gambar 1 Volume Total Lalu lintas Jam-Jaman Simpang tiga Straat A
Hari Sabtu, 01 November 2014
2. Kapasitas
Berdasarkan perkalian kapasitas dasar (C0) pada kondisi ideal simpang dengan faktor-
faktor koreksi (F), diperoleh kapasitas disimpang sebesar 6.705 smp/jam.
3. Derajat Kejenuhan (DS)

Derajat kejenuhan dihitung dengan persamaan 2.14:
DS = 0,418 < 0,850
4. Tundaan (D)
Tundaan lalu lintas disimpang (DTi) 4,40 det/smp, (DMA) 3,27 det/smp, (DMI) 5,48
det/smp, tundaan geometri simpang (DG) 5,16 det/smp sehingga dapat dilihat besar
tundaan simpang (D) 9,56 det/smp.
5. Peluang Antrian
Pada analisis peluang antrian yang ditentukan dari hubungan empiris peluang antrian
terhadap derajat kejenuhan didapat nilai batas maksimum 19,75 % dan batas minimum
8,14 %.
Hasil analisis perhitungan dengan metode MKJI 1997 menghasilkan kinerja simpang berupa
kapasitas simpang, volume total simpang, tundaan simpang dan tingkat pelayanan simpang
seperti terlihat pada tabel.8 di bawah.
225
Tabel 8. Rekapitulasi Hasil Analisis Simpang Tiga Straat A
Analisis dan Pembahasan Ruas Jalan Ahmad Yani

1. Volume Lalu lintas
Rekapan perhitungan jam-jaman volume Lalu lintas dapat dilihat pada Gambar 2
yang ditampilkan adalah total volume lalu lintas yang terbesar (maksimum) dari ruas
jalan A. Yani:
Gambar 2 Volume Lalu lintas Jam-Jaman di Jalan A. Yani
Total volume (Q) yang dipakai merupakan volume kendaraan terbesar dalam dalam 1 jam
puncak pada ruas Jalan Ahmad Yani dengan nilai Q sebesar 1.646 SMP/jam.
2. Hambatan Samping
Berdasarkan hasil pengamatan visual terlihat bahwa kelas hambatan samping ruas jalan
Ahmad Yani berada pada level sedang (M). Jalan Ahmad Yani melewati daerah komersil
dengan beberapa toko di sisi jalan, dimana toko-toko tersebut umumnya memiliki tempat
parkir Off Street.
3. Kecepatan Arus Bebas

Perhitungan untuk kecepatan arus bebas terhadap ruas Jalan Ahmad Yani menggunakan
rumus 17 diperoleh sebagai berikut:
FV = ( Fvo + FVw) x FFVsf x FFVcs
FV = (55 + 4) x 0.89 x 0.95
FV = 49,88 km/jam
4. Kapasitas Jalan
226
Setelah menentukan koefisien nilai-nilai pada rumus 18 kapasitas jalan, maka kapasitas
jalan Ahmad Yani adalah:
C = CO x FCw x FCsp x FCsf x FCcs
C = 3.300 x 1,21 x 1.00 x 0,88 x 0,94
C = 3.303 SMP/jam
5. Perhitungan Derajat Kejenuhan

Nilai dejat kejenuhan dihitung dengan rumus DS = Q/C
DS =Q/C
DS = 1.646/3.303
DS = 0,49
Nilai DS pada jalan Ahmad Yani berdasarkan hasil perhitungan adalah 0,49. Berdasarkan
nilai DS yang didapat Tingkat pelayanan C menunjukan bahwa pada pada ruas jalan
Ahmad Yani arus stabil tetapi pergerakan dan kecepatan kendaraan dikendalikan oleh
volume Lalu lintas.
6. Kecepatan Aktual Kendaraan

Kecepatan aktual yang dapat ditempuh pada ruas jalan Ahmad Yani adalah 48 km/jam.
7. Waktu Tempuh
Waktu tempuh kendaraan pada ruas Jalan Ahmad Yani diperoleh dari hasil bagi panjang
segmen jalan dengan kecepatan aktual.
= 0.525 km / 48 km/jam
= 39,40 detik
8. Rekapitulasi Hasil Analisis dari ruas –ruas Jalan A. Yani, Jalan Sumba, Jalan Flores
Berikut rekapitulasi hasil analisis volume (Q), kecepatan arus bebas (FV), kapasitas
(C ), derajat kejenuhan (DS), kecepatan actual (VLV), waktu tempuh (TT) dari ruas
Jalan A. Yani, Jalan Sumba, Jalan Flores.
Tabel 9. Rekapitulasi Hasil Analisis ruas jalan
227
KESIMPULAN DAN SARAN

Kesimpulan
Berdasarkan hasil penelitian analisis arus lalu lintas di Simpang Tiga Straat A dan ruas-ruas jalan
yang dipengaruhi dapat disimpulkan sebagai berikut:
1. Dengan besar volume total simpang 2.803 smp/jam, kapasitas 6.705 smp/jam, tundaan
simpang 9,56 det/smp, nilai derajat kejenuhan (DS) di simpang = 0,418, tundaan lalu lintas
4,40 det/smp, serta peluang antrian sebesar 8,14 % - 19,75 %. Mengindikasikan simpang
menurut MKJI kondisi ini dikategorikan B dengan arus stabil, lalu lintas sedang, kecepatan
mulai dibatasi, hambatan belum mempengaruhi, pengemudi masih punya cukup kebebasan
memilih kecepatan.
2. Setelah diterapkannya rekayasa lalu lintas sistem satu arah kinerja ruas - ruas jalan yang
dipengaruhi:
a. Ruas jalan A. Yani berada pada tingkat pelayanan C, Dengan karakteristik:
1) Volume kendaraan maksimum pada jam puncaknya adalah 1.646 smp/jam.
2) Kejadian hambatan samping yang terjadi di jalan A. Yani tergolong Sedang
(kategori M).
3) Kecepatan aktual pada saat terjadi volume puncak dan hambatan samping terbesar
adalah sebesar 48 km/jam.
4) Nilai Derajat Kejenuhan (DS) pada jalan A. Yani berdasarkan hasil analisa adalah
0,49. Dari nilai tersebut maka diperoleh Tingkat Pelayanan pada ruas jalan Ahmad
Yani adalah pada tingkat Pelayanan C.
b. Ruas jalan Sumba berada pada tingkat pelayanan C, Dengan karakteristik:
1) Volume kendaraan maksimum pada jam puncaknya adalah 1.509 smp/jam.
2) Kejadian hambatan samping yang terjadi di Jalan Sumba tergolong Sedang (kategori
M).
4) Nilai Derajat Kejenuhan (DS) pada jalan Sumba berdasarkan hasil analisa adalah
0,48. Dari nilai Derajat Kejenuhan tersebut maka diperoleh Tingkat Pelayanan Jalan
pada ruas Jalan Sumba adalah pada tingkat Pelayanan C
c. Ruas Jalan Flores berada pada tingkat pelayanan C, Dengan karakteristik:
1) Volume kendaraan maksimum pada jam puncaknya adalah 1.342,9 smp/jam.
2) Kejadian hambatan samping yang terjadi di Jalan Flores tergolong Sedang (kategori
M).
4) Nilai Derajat Kejenuhan (DS) pada Jalan Flores berdasarkan hasil analisa adalah
0,48. Dari nilai Derajat Kejenuhan tersebut maka diperoleh Tingkat Pelayanan Jalan
pada ruas Jalan Flores adalah pada tingkat Pelayanan C.
Saran
Beberapa saran sehubungan dengan hasil analisis terhadap tingkat pelayanan simpang dan ruas
adalah sebagai berikut:
1. Hasil penelitian ini bisa diterapkan oleh Pemerintah Kota Kupang dan instansi - instansi
terkait, karena dari hasil analisis didapat kinerja simpang dan ruas jalan yang lebih baik.
2. Perlu dibuatkannya pemberhentian sementara bagi mobil penumpang, agar mobil-mobil
tersebut tidak berhenti sembarangan/ sesukanya dibadan jalan, karena kondisi tersebut
sangat mengganggu arus lalu lintas.
3. Direkomendasikan juga kelas jalan dan geometri untuk ruas Jalan Flores disesuaikan dengan
ruas Jalan A. Yani dan ruas Jalan Sumba. Peningkatan kualitas jalan dengan cara pelebaran
228
jalan, sangat dibutuhkan guna memperbesar lebar efektif jalan tersebut. Hal ini berdampak
pada kapasitas dan tingkat keselamatan jalan tersebut.
DAFTAR PUSTAKA
Anonim. 1997. Manual Kapasitas Jalan Indonesia (MKJI), Swe Road in Association
With PT Bina Karya
Anonim. 2000. Highway Capacity Manual (TRB 2000), Tranportation National Board
Edward K. Morlok. 1978. Pengantar Teknik dan Perencanaan
Transportasi. Penerbit Erlangga.
Frans Christine D, 2014. Analisis Kinerja Simpang Tiga Bersinyal Straat A, Universitas
Nusa Cendana, Kupang
Khisty, C. J dan Kent L.B. 2005. Dasar-dasar Rekayasa Transportasi : Penerbit Erlangga
Miro, Fidel (2005). Perencanaan Trasnportasi : Untuk Mahasiswa, Perencana dan
Praktisi. Erlangga, Jakarta
Oglesby, Hicks. 1999. Teknik Jalan Raya Jilid 1, Erlangga, Jakarta
Peraturan Pemerintah No. 43 tahun 1993, Sarana dan Prasarana Jalan, Kementrian
Perhubungan Indonesia
Peraturan Pemerintah No. 26 tahun 1985, Klasifikasi Jalan, Kementrian Perhubungan
Indonesia
Risdiyanto, (2008).Perbandingan Tundaan Simpang Bersinyal Dengan Metode MKJI
1997 Dan Metode Survei Lapangan . Universitas Janabadra. Yogyakarta
(Jurnal)
Takoy Diana, 2010. Analisis Tingkat Pelayanan Jalan Arteri Di Kota Kupang (Studi
Kasus Jalan Siliwangi Dan Jalan Ahmad Yani), Universitas Nusa Cendana,
Kupang
Undang-Undang No. 14 tahun 1992, Lalu Lintas dan Angkutan Jalan, Kementrian
Perhubungan Indonesia
229
230
Version of Record: https://www.sciencedirect.com/science/article/pii/S1367578819300446
Manuscript_e3963856058ab91a75c58b3dfe2045fe
Ecological traffic management: a review of the

modeling and control strategies for improving
environmental sustainability of road transportation
Bassel Othmana,b,∗, Giovanni De Nunzioa , Domenico Di Domenicoa , Carlos

Canudas-de-Witb
a IFP Energies nouvelles, Rond-point de l’échangeur de Solaize, BP 3, 69360 Solaize, France
b Univ. Grenoble Alpes, CNRS, Inria, Grenoble INP, GIPSA-lab, 38000 Grenoble, France
Abstract
As road transportation energy use and environmental impact are globally rising
at an alarming pace, authorities seek in research and technological advancement
innovative solutions to increase road traffic sustainability. The unclear and par-
tial correlation between road congestion and environmental impact is promoting
new research directions in traffic management. This paper aims to review the
existing modeling approaches to accurately represent traffic behavior and the
associated energy consumption and pollutant emissions. The review then covers
the transportation problems and control strategies that address directly environ-
mental performance criteria, especially in urban networks. A discussion on the
advantages of the different methods and on the future outlook for the eco-traffic
management completes the proposed survey.
Keywords: traffic modeling, traffic management, energy efficiency, pollutant
emissions, optimization
1. Introduction
While energy-related air pollution is considered today one of the primary

premature death causes (World Health Organization, 2016), the global carbon
dioxide (CO2 ) emissions are on a rising trend destined to grow well above the
5 levels imposed by the international climate goals (International Energy Agency,
2018). Population surge and economic growth of the developing countries have
been identified as the main causes of the drastic increase of energy demand and
pollutant emissions in all sectors (International Energy Agency, 2018).
∗ Corresponding author.
E-mail addresses: bassel.othman@ifpen.fr (B. Othman), giovanni.de-nunzio@ifpen.fr

(G. De Nunzio), carlos.canudas-de-wit@gipsa-lab.fr (C. Canudas de Wit),
domenico.didomenico@ifpen.fr (D. Di Domenico).
Preprint submitted to Annual Reviews in Control June 28, 2019
© 2019 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
The worldwide transportation sector alone accounts for 55% of the total liq-
10 uid fuels consumption and, with the increasing travel demand, this share is not
expected to decrease for the next two decades (U.S. Energy Information Admin-
istration, 2017a). In the member countries of the Organisation for Economic
Co-operation and Development (OECD), projections show that the improved
energy efficiency in transportation may lead to a net decline of about 2% in
15 energy use until 2040, thus outpacing the predicted increase of vehicle-miles
traveled (VMT). However, in OECD-Europe, transportation still represents the
biggest source of carbon emissions (Transport & Environment, 2018), contribut-
ing by about 25% to the total CO2 emissions, with cars and vans representing
more than two thirds of this share (Mandl and Pinterits, 2018). The situation
20 is even more alarming in non-OECD countries, where the transportation energy
demand is expected to rise by 64% until 2040, implying an increase of about
15% of energy-related CO2 emissions (U.S. Energy Information Administration,
2017a).
Therefore, a lot of attention has been drawn worldwide to finding the most ef-
25 fective measures to help reduce the current contribution to greenhouse gas emis-
sions from transportation. Governments, practitioners and researchers seem to
agree on the fact that a combination of short-term and long-term strategies must
be adopted. In the short-term, policies and regulations encouraging changes in
behavior and travel habits represent a key lever. Attractiveness of alternative
30 means of transportation should be enhanced, a shift to less polluting transport
modes should be promoted, and a change in purchasing habits favoring smaller
and more energy-efficient cars should be encouraged (Chapman, 2007). In the
long-term, the widespread adoption of innovative technological solutions such as
electrification, connectivity and automation are expected to enable a significant
35 shift in the future of personal transportation and mobility. The way for such
a technological transformation of mobility is already being paved thanks to the
diffusion of connected and automated vehicles (CAVs), multi-vehicle (V2V) and
vehicle-infrastructure (V2I) cooperation and communication networks, in- and
over-roadway sensors, cloud-computing capabilities, etc. (Guanetti et al., 2018).
40 However, the potential energy benefits of these technologies remain uncer-
tain, mostly because of the high level of non-linear dependence between different
aspects of an automated transportation system operating with conventional ve-
hicles, as well as possible side-effects of automation (U.S. Energy Information
Administration, 2017b). Among the features enabled by the aforementioned
45 technologies that promise to increase energy efficiency and reduce pollutant
emissions of transportation, it is worth mentioning eco-driving, eco-routing,
platooning, roadway throughput optimization, powertrain electrification, vehi-
cle down-sizing, parking search time reduction, ride-sharing. On the other hand,
as for the side-effects that may endanger energy efficiency and emission reduc-
50 tion, it is likely that technology may increase traffic congestion as a consequence
of an increased access to mobility, increase travel speeds as a consequence of en-
hanced safety, increase commute distances as an effect of increased comfort and
reduced travel costs, etc. (U.S. Energy Information Administration, 2017b).
From a single-vehicle efficiency perspective, research suggests that lightweight,
2
55 low-speed, autonomous vehicles have the potential to achieve fuel economies an
order of magnitude higher than current cars (U.S. Energy Information Admin-
istration, 2017b). However, at system-wide level, current estimates suggest that
the total energy consumption impacts can range from a 90% decrease to a 200%
increase in fuel consumption as compared to a projected 2050 baseline energy
60 (Brown et al., 2014).
Such a large variability in the possible outcome of the adoption of the new
vehicular and traffic technologies makes it somewhat difficult to focus and pri-
oritize the research efforts to increase energy efficiency of mobility. Nowadays,
the general trend in research and policy seems to aim to reduce CO2 emissions
65 by pushing for more efficient vehicles and reducing VMT. This is based on a
generally accepted paradigm that congestion mitigation programs should reduce
CO2 emissions. However, it is difficult to prove a clear direct proportionality
between congestion and CO2 emissions (Fiori et al., 2018). The most reliable
approach to improve energy efficiency and reduce pollutant emissions in the de-
70 sign of a traffic regulation measure consists in directly considering these aspects
as decision and optimization criteria. Therefore, interest in transportation reg-
ulation problems with explicit environmental considerations is growing (Wang
et al., 2018; Vreeswijk et al., 2013).
This paper surveys the existing scientific literature on energy consumption
75 and emission models, as well as road transportation problems directly address-
ing the issue of energy consumption and pollutant emissions reduction. Such
problems can be tackled at different levels depending on the granularity and
the object of the control action. At vehicle level, the energy-efficient control
strategies typically act on single vehicles or groups of cooperating vehicles by
80 modifying their individual speed profiles or route choices. At traffic level, the
control strategies aim to influence the vehicular flow as a whole by acting on
the typical flow regulation actuators, such as traffic lights, speed limits, etc.
The adopted categorization in terms of modeling and control approaches both
at vehicle and traffic level for the general problem of reducing environmental
85 impact of road transportation is illustrated in Fig. 1.
The contributions of this paper are summarized as follows:
• A comprehensive literature review of the existing energy consumption and
pollutant emissions models is provided. The review distinguishes between
data and physics-based models and discusses their adaptation for usage
90 with both single-vehicles and traffic flow.
• An overview of the existing vehicle and traffic control strategies to im-
prove energy and environmental efficiency of transportation is given. The
review focuses on the control techniques that explicitly address energy con-
sumption and emissions. The connection and interaction between traffic
95 congestion and energy efficiency is also discussed.
• As an outcome of this review, research gaps in the current state of the art
have been identified and discussed in order to inspire future works in this
field.
3
Single vehicle Emission/energy
Kinematics Control
models
Speed Data-driven Eco-driving

measurements
Microscopic Physical Eco-routing

traffic models
Emission/energy
Traffic Kinematics Control
models
Speed Traffic
Data-driven lights
measurements
Speed
Fluid-based Physical limits
models
Dynamic
routing
Automated
vehicles
Figure 1: Diagram of the global approach for energy consumption and emissions modeling
and control for single vehicles and traffic flow.
The body of the paper is organized as follows. Section II presents the energy
100 consumption and emission models for the single vehicle with a brief discussion of
how the vehicle kinematics can be obtained. Analogously, Section III introduces
the modeling approaches to describe traffic kinematics, with a particular focus
on the most popular fluid-dynamics traffic models, as well as the energy con-
sumption and emission models for vehicular flow. The energy-optimal control
105 strategies for single vehicles are presented in Section IV, while the transporta-
tion problems dealing with traffic energy efficiency are reviewed in Section V.
Finally, Section VI contains concluding remarks and discussion on the current
research gaps and future outlooks.
2. Emission and energy consumption models for single vehicles
110 Different models estimating emissions and energy consumption rate (Jy ) of
a vehicle as a function of its parameters and operation variables (u) have been
investigated in the past. This section presents the data-driven and the physical
modeling approaches employed to estimate Jy .
In the proposed formalization, Jy refers to the prediction of the rate of y,
which can be calculated per distance traveled by the vehicle (Jyspat ) or per time
unit (Jytemp ), depending on the modeling method. y corresponds either to the
emission of a pollutant (CO, N Ox , HC, ...) or the energy consumption (fuel or
4
Microscopic
emission/energy models
Data-driven Physical modeling

approach approach
Look-up table Regression Deterministic Probabilistic

models models methods methods
Figure 2: Classification of emission and energy consumption models for single vehicles.
electricity consumption, depending on the vehicle powertrain considered):

y ∈ {fuel or electricity consumption, emission of CO, N Ox , HC, ...}
Such emission and energy consumption models are said microscopic because
115 they consider each vehicle individually. They can be described as
Jy = f (u) (1)
where f is a function that relates the model inputs to the output.
The function f can be constructed in different ways. The different ap-
proaches detailed in this section to estimate emissions and energy consumption
are classified as illustrated in Fig. 2.
120 The first step to determine the emission and energy consumption rates of
a vehicle is to determine its operation variables (e.g. speed, acceleration). A
solution is to obtain these data by sensors. For example, Thibault et al. (2016)
propose to use smartphone devices and their embedded sensors to get the posi-
tion and speed of vehicles. Treiber and Kesting (2013b) present a methodology
125 to express the operation variables of vehicles from trajectory and floating-car
data.
They can also be determined through simulation using a microscopic traffic
model, which reproduces the movement of each vehicle individually. Some com-
plete overviews of microscopic traffic models can be found in van Wageningen-
130 Kessels et al. (2015); Ferrara et al. (2018b); Hoogendoorn and Bovy (2001).
These approaches are mainly based on the car-following principle (e.g. safe-
distance models, stimulus-response models, action point models). For example,
the optimal velocity car-following model expresses the acceleration of each ve-
hicle as
ve − v
a= (2)
τ
135 where the optimal speed, ve , depends on the distance with the vehicle upstream,
and τ is the driver reaction time.
The use of a microscopic traffic model, especially in order to estimate emis-
sions and energy consumption, requires a precise calibration of model param-
eters. Jie et al. (2013) present a methodology to perform such a parameters
5
Symbol Description of the variable
a Vehicle acceleration [m.s−2 ]
A Cross-sectional area [m2 ]
by Stoichiometric CPF (Catalyst Pass Fraction)
coefficient [s.kg−1 ]
cy Enrichment CPF coefficient [s.kg−1 ]
Ca Aerodynamic drag coefficient [-]
Cd Reynolds coefficient [-]
Ce Engine friction factor [J.rev−1 .m−3 ]
Cr Rolling resistance coefficient [-]
COC Center of combustion (50% energy conversion, from
Top Dead Center) [crank angle degree]
CP Fy Catalyst pass fraction of y [-]
d Mass density of air [kg.m−3 ]
D Engine displacement [m3 ]
g Gravitational constant [m.s−2 ]
LHVfuel Fuel lower heating value [J.kg−1 ]
mcyl In-cylinder air mass per stroke and displaced
volume [kg.m−3 .sr−1 ]
mO2 In-cylinder oxygen mass per stroke and displaced
M Vehicle mass [kg]
n Engine speed [revolutions per minute]
Pacc Engine power demand associated with accessories [W]
RBGR In-cylinder burnt gas ratio [-]
v Vehicle speed [m.s−1 ]
y Maximum catalyst efficiency of y [-]
α Road grade angle [rad]
λ Ratio between the air/fuel ratio at stochiometry and
the commanded air/fuel ratio [-]
ηbatt Battery efficiency [-]
ηeng Engine efficiency [-]
ηtf Efficiency of the transmission and final drive [-]
Table 1: Parameters and operation variables of vehicles used in the emission and energy
consumption models.
140 calibration and emphasize on its benefits in terms of speed and acceleration
estimation.
The second step to determine the emission and energy consumption rates
of a vehicle is to use a microscopic emission and energy consumption model
whose inputs are the vehicle operating variables and parameters, summarized
145 in Table 1. This step is presented in detail in the following sections.
6
2.1. Data-driven methods
Emission and energy consumption rates can be calculated using data-driven
approaches. These can be either based on look-up table models or regression
models.
150 2.1.1. Look-up table models

One old common approach to estimate emission and fuel consumption rates
consists in performing chassis dynamometer tests and recording the emissions
and fuel consumption in a look-up table, also called emission matrix. Usually,
such look-up tables provide Jy from speed and acceleration (Post et al., 1984;
155 Sturm et al., 1998) for a given set of vehicle parameters. These reference emis-
sion look-up tables can be used later to instantly estimate emissions and fuel
consumption.
Although this method is easy to use, usually the available matrices are
sparse, due to measurement difficulties. Moreover, empirical matrix-based pre-
160 diction concerns only steady-state emissions, and not transient operations (Scora
and Barth, 2006). Finally, this method is sensitive to the driving cycle and the
quality of on-line measurements. This may lead to large errors.
Another possibility is to determine emission and fuel consumption rates as a
function of the vehicle position. Andersen et al. (2013) propose to associate to
165 each road a corresponding fuel consumption, based on average measurements.
The amount of fuel consumed by a vehicle during a trip is therefore simply
approximated by the sum of the average fuel consumption associated with the
corresponding roads. This approach is very simple but it cannot distinguish
between different types of vehicle, as they are all mixed in the same computed
170 average value. Also, it cannot reflect the evolution of emissions and fuel con-
sumption in case of traffic congestion.
2.1.2. Regression models

Emissions and energy consumption of a single vehicle can also be predicted
on a second-by-second basis by using data-based models, such as regression
175 techniques or neural networks. The inputs of these models can typically be
the speed, acceleration or power demand, and the outputs are the emission or
energy consumption rates prediction.
Regression techniques and neural networks for emission and energy consump-
tion modeling both use the collected data in order to train a model that mimics
180 these data. In regression techniques, it is necessary to identify the model pa-
rameters by curve fitting, while in neural networks the weight of the connections
between neurons is to be identified.
The use of neural networks to estimate emissions and energy consumption
is motivated by the heavy nonlinearity of emissions. There is also a need of
185 high computational efficiency in order to be compatible with second-by-second
microscopic traffic models. Such neural network frameworks can be found in
Ahn (1998); Obodeh and Ajuwa (2009); Jafarmadar (2015); Xu et al. (2017).
Ahn (1998) presents non-linear multiple regression models constructed with
quadratic and cubic speed-acceleration terms. The data used to determine the
7
190 coefficients of these models for a given type of vehicle is obtained from dy-
namometer emission tests, based on the New European Drive Cycle (NEDC)
(Ahn et al., 2004). It is also desirable to use data from vehicles in real urban
traffic situations, when available (Panis et al., 2006). In fact, it is important to
note that emission levels obtained from dynamometer tests can be much lower
195 than those produced in real traffic (Pelkmans and Debal, 2006). For example,
a criticism against the NEDC is that its acceleration profile is very smooth and
not sufficiently realistic (Andre and Pronello, 1997).
Based on this technique, the VT-micro model can be formulated in matrix
form (Zegeye et al., 2013) as
ln(Jytemp ) = ~v My~a (3)
200 where My denotes the regression coefficients matrix of y for the type of vehicle
under consideration, ~v and ~a are respectively the speed and acceleration vectors
defined as
~v = 1, v, v 2 , v 3

T (4)
~a = 1, a, a2 , a3

Note that the VT-micro model can also be expressed with a regression co-
efficients matrix for positive accelerations, and another matrix for negative ac-
205 celerations, depending on the data used to calibrate the model (Alsabaan et al.,
2012). VT-micro estimates emissions and energy consumption from instanta-
neous speed and acceleration, i.e. measured at the present time. Qi et al. (2004)
formulate a regression model, named POLY, which also takes into account the
past accelerations and the road grade angle. The model reads
Jytemp = β0 + β1 v(k) + β2 v 2 (k) + β3 v 3 (k)
+β4 T acc (k) + β5 T dec (k) (5)
+β6 ga (k) + ... + β15 ga (k − 9) + β16 v(k)ga (k)

210 where β0 to β16 are the parameters determined by least-square method for
one type of vehicle, T acc (k) and T dec (k) are respectively the acceleration and
deceleration duration since their inception up to the current time step k. At
each time step, at least one of them is zero. To consider the grade angle α, the
function ga is defined as follows
" #
α(k)
ga (k) = a(k) + g p (6)
1 + α2 (k)
215 POLY is an accurate emission model. However, it may underestimate emis-
sions of higher emitting vehicles as it is built from average measured data (Qi
et al., 2004).
While data-driven models can be developed quickly without prior knowledge
on the vehicle or roads, they usually lack a clear physical interpretation and
220 might be too coarse. They may also over-fit the calibration data if the number
of variables considered is too large.
8
2.2. Physical modeling approach
An alternative method for estimating emissions and energy consumption is
to employ a physical approach that leads to model parameters with physical
225 meaning. Two types of models can be distinguished, the deterministic and
probabilistic models, that are both described in the following sections.
2.2.1. Deterministic methods

The emission and energy consumption rates can be determined from the
power engine demand P , which can be calculated using the following vehicle
230 longitudinal dynamics, as in Sciarretta et al. (2015)

M a = Ftrac − Fb − Fres
(7)
a = dv
dt
where Ftrac is the traction force transmitted by the powertrain to the wheels,
Fb is the mechanical brake force and Fres is the resistance force that can be
calculated as follows
1
Fres = M g sin α + M gCr + dv 2 ACa (8)
2
The total tractive power of the vehicle is denoted Ptrac and is given by
Ptrac = Ftrac v (9)
235 (7) to (9) lead to

d
Ptrac = M v (a + g sin α) + v M gCr + v 2 ACa + Fb v (10)
2
Finally, the power engine demand P can be calculated as follows
Ptrac
P = + Pacc (11)
ηtf
Once the power demand is known, Post et al. (1984) propose to estimate the
emission and energy consumption rates as follows

ay + by P , if P ≥ 0
Jytemp = (12)
ay , if P < 0
ay and by are the regression coefficients determined for a given y and vehicle
240 type. When y represents the fuel consumption, ay can be approximated by the
following linear function
ay = γ × D (13)
where γ is a constant.
Barth et al. (1996) propose to replace the regression coefficients by physical
parameters and operation variables to approximate the fuel use rate. The model
245 is defined as
temp P
Jf uel LHVf uel ≈ λ Ce nD + (14)
ηeng
9
where the engine friction factor Ce is the energy used at zero-power output to
overcome engine friction.
An et al. (1997) then propose to calculate the pollutant emission rates as
follows
dy
Jytemp = Jftemp
uel CP Fy (15)
d(f uel)
250 where y refers here only to emissions, d(fdy
uel) corresponds to the grams of engine-
out emissions per gram of fuel consumed for pollutant y, and the catalyst pass
fraction CP Fy can be modeled as

1
CP Fy = 1 − y exp −by − cy 1 − Jftemp
uel (16)
λ
The Comprehensive Modal Emissions Model (CMEM) is based on (15)
(Scora and Barth, 2006). It considers different categories of vehicle and different
255 modes of operation (idling, cruising, acceleration, and deceleration). Emission
and fuel consumption rates are calculated as a function of the vehicle fleet com-
position (vehicle categorization based on model year, weight, etc.), operation
variables and model-calibrated parameters. The structure of the model is shown
in Fig. 3. The CMEM predicts emissions well, but may underestimate them
260 for high-emitting vehicles because the model is based on the average data of
300 vehicles (including about 30 high emitters) measured during dynamometer
tests, along different driving cycles (Rakha et al., 2003).
Figure 3: Structure of the CMEM (Scora and Barth, 2006) [Published with permission of the
Center for Environmental Research and Technology].
Another model proposed by Gärtner et al. (2004) estimates emissions from

fuel consumption at engine mechanics level. The model relies on the first Law of
10
265 Thermodynamics and chemical kinetic reaction rate considerations. For N Ox
emissions, the model reads

d(N Ox )
log = a0 + a1 COC + a2 mcyl + a3 mO2 (17)
d(f uel)
where d(N Ox )
d(f uel) is the mass of nitrogen oxides emitted per mass of fuel consumed,
and a0 , a1 , a2 , a3 are model coefficients.
A simplified version of (17) is proposed by Thibault et al. (2016). The
270 authors propose to express log(JN Ox ) as a linear function of the in-cylinder
burnt gas ratio RBGR , as follows

d(N Ox )
log = a4 + a5 RBGR (18)
d(f uel)
where a4 , a5 are model coefficients.

RBGR is expressed as a function of the engine speed and the engine torque
for a given type of vehicle, based on the data from the NEDC. The engine con-
275 ditions are physically determined from the speed of the vehicle and its constant
parameters.
The same approach can be considered to estimate the emissions of other
pollutants.
2.2.2. Probabilistic methods

280 The previous models estimate emissions and energy consumption as a func-
tion of real vehicle operation variables (e.g. speed and acceleration, power de-
mand, engine mechanics, etc.).
However, these data are not always available. One may obtain the velocity
through microscopic traffic model simulation. But such models can be difficult
285 to implement, especially on a large spatial scale with a lack of precise knowl-
edge about the traffic situation, and can lead to unrealistically smooth velocity
profiles. Hence, probabilistic models, based on random velocity disturbances,
have been proposed in the literature.
The general idea of the random velocity disturbances approach is to run the
290 emission and energy consumption models while replacing, for a given route, the
actual speed of the vehicle by an approximate second-by-second speed profile
built from a deterministic and a stochastic component as
ṽ = v̄ + Θ (19)
where v̄ is the average traffic speed estimated from the road attributes provided
by a geographical information system (e.g. speed limit, traffic signs, road grade,
295 etc.) and Θ is a random variation in velocity for the subject vehicle.
One possibility is to consider a spatial distribution of speed or acceleration
based on driving cycles or statistical distributions (Burghout, 2004).
Karbowski et al. (2014) combine Markov chains with deterministic route
attributes to generate the speed profile. In this model, Θ is adjusted according
11
300 to
P (X(k + 1) = Xi |X(k) = Xj ) = MT P (i, j) (20)
where X(k) = [v(k) a(k)]T is the state vector of the vehicle at time step k, and
the transition probability matrix MT P is built from real data.
Another probabilistic model is the Motor Vehicle Emission Simulator (MOVES),
presented by Wu et al. (2014). The aim of this method is to make the veloc-
305 ity trajectory more realistic. Thus, it is assumed that vehicle detector stations
provide an estimation of v̄. The random variation in velocity is defined as
Θ(k) = ṽ(k − 1) + a(k − 1) − v̄(k) (21)
A procedure to determine the acceleration a is presented in Wu et al. (2014).

Probabilistic approaches are a solution in case of lack of information about
the vehicle dynamics. By construction, they are less accurate than models
310 based on the actual speed, but can be effectively used to estimate emissions and
energy consumption (Kubička et al., 2016). To improve these methods, traffic
prediction models could be integrated to determine v̄ (cf. Section 3.1.2).
Note that the variability of certain unobserved parameters between vehicles
(e.g. temperature, Reid vapor pressure) can affect the emissions and energy con-
315 sumption. These issues can be addressed by introducing probabilistic correction
factors (Frey and Zheng, 2002).
3. Emission and energy consumption models for traffic vehicular flows
The emission and energy consumption models presented in Section 2 are

microscopic. They estimate emissions and energy consumption based on the
320 instantaneous operating variables of individual vehicles, that can be obtained
through microscopic traffic models. But on a network scale, they have the known
disadvantages of high computational load, as their computation time increases
sharply with the number of vehicles. The instantaneous operating variables can
also be measured, but the data for so many vehicles are very difficult to obtain
325 and process.
For large scale control purposes it is necessary to develop macroscopic models
that use aggregate network or link-based data to estimate global emissions and
energy consumption. These models are more coarse but also simpler to use and
allow for faster computation. They are based on the traffic variables presented
330 in Table 2.
In this section, we first review how to determine the traffic kinematics, then
we present different emission and energy consumption models that can be set
up.
3.1. Traffic kinematics

335 To determine the traffic kinematics, it is possible to measure the average
speed of vehicles, or to use a traffic model based on fluid dynamics.
12
i Cell index [-]
k Discrete Time index [-]
L Length of link [m]
Ni (k) Number of vehicles in cell i at time step k [veh]
v Traffic speed [m.s−1 ]
v̄ Average traffic speed [m.s−1 ]
vmax Maximum speed, forward wave speed [m.s−1 ]
δ Discrete-time step size [s]
δL Discrete-space cell length [m]
ρ Vehicle density [veh.m−1 ]
ρcr Critical density, i.e. associated with the maximum flow [veh.m−1 ]
ρM Maximum possible vehicle density [veh.m−1 ]
ϕ Traffic flow [veh.s−1 ]
ϕi (k) Traffic flow entering cell i during
time interval [kδ, (k + 1)δ] [veh.s−1 ]
ϕM Maximum possible traffic flow [veh.s−1 ]
w Backward wave speed [m.s−1 ]
Table 2: Traffic variables and indices used in emission and energy consumption models.
3.1.1. Average speed

The average speed of the traffic on each link i is defined as
T N (i,k)
1 X 1 X
v̄(i) = v j (k) (22)
T N (i, k) j=1
k=1
where T is the number of time steps on which the average speed is performed,
340 N (i, k) is the number of vehicles on link i at time step k and v j (k) denotes the
speed of vehicle j at time step k. In the following, v̄(i) is referred to as v̄ for
simplicity.
The average speed can be provided using fixed sensors or Floating Car Data
(FCD) methods, like the smartphone devices of the drivers for example. Sim-
345 ilarly, the number of vehicles N (i, k) can be provided by induction loops or
cameras.
3.1.2. Fluid-based models

The traffic kinematics can also be determined through dynamic fluid-based
traffic models that describe the evolution of the traffic in the network as a fluid
350 in a pipe. Some overviews presenting this kind of model can be found in Ferrara
et al. (2018a,c); van Wageningen-Kessels et al. (2015); Hoogendoorn and Bovy
(2001).
This approach provides the traffic variables, i.e. ρ(x, t), v(x, t), and ϕ(x, t),
at given position x and time t. It considers the traffic speed as a function of the
355 position x and the time t. Therefore, unlike the average speed approach, these
13
models reflect the speed differences along links and provide a dynamic traffic
speed.
Some macroscopic traffic models are reviewed in the following. They are all
based on the following conservation law
∂ ∂
ρ(x, t) + (ρ(x, t)v(x, t)) = 0 (23)
∂t ∂x
360 Some of these models are continuous and others are spatially and temporally
discretized. A distinction is made between first and higher order models.
– First order models
• Lighthill-Whitham-Richard model
Lighthill and Whitham (1955) and Richards (1956) assume that v depends
365 only on ρ. Hence, the flow can be expressed as a function of only ρ as
ϕ = ρv(ρ) = Φ(ρ) (24)
The conservation law presented in (23) can then be expressed as
∂ ∂
ρ+ Φ(ρ) = 0 (25)
∂t ∂x
where Φ is a strictly concave C 1 function defined on [0, ρM ] and satisfying Φ(0) =
Φ(ρM ) = 0.
The relationship ϕ = Φ(ρ) is called the fundamental diagram. The most
370 common fundamental diagrams are listed in Table 3 (Garavello et al., 2016), in
which v0 is a positive constant.
p
Greenshields et al. (1935) Φ(ρ) = ρvmax 1 − ρρM , p∈N
M
ρ
Greenberg (1959) Φ(ρ) = ρv0 ln ρ

Underwood Φ(ρ) = ρvmax exp − ρρM

California Φ(ρ) = ρv0 ρ1 − ρ1M
M
Trapezoidal (Daganzo, 1994) Φ(ρ) = min ρv max , ϕ , M (ρM − ρ)w
Triangular (Newell, 1993) Φ(ρ) = min ρvmax , (ρ − ρ)w
Table 3: List of most common fundamental diagrams
• Cell transmission model

Daganzo (1994) proposes the cell transmission model (CTM) which is a
temporally and spatially discretized version of the LWR model based on the
375 triangular or the trapezoidal fundamental diagram. The model is defined as
 ρi (k + 1) = ρi (k) + δδL (ϕi (k) − ϕi+1 (k))


(26)
ϕi (k) = min ρi−1 (k)vmax , ϕM , w(ρM − ρi (k))

14
• Variable-length model
In order to depict density evolution and track the congestion front, Canudas-
de-Wit (2011) proposes the variable-length model (VLM) for highway traffic
modeling. Illustrations are given on a closed ring road and on an urban road
380 with traffic lights in Canudas-de-Wit and Ferrara (2018).
The VLM is also a discrete version of the LWR model based on the triangular
fundamental diagram. The idea is to model any road section with only two
lumped cells that are variable in length: an upstream cell in free flow and
a downstream congested cell. Consider a road section of length L, then the
385 length of the free and the congested cells will respectively be L − l and l.
The main advantage of the VLM is that it is based on only three state vari-
ables: density in the upstream free cell ρf , density in the downstream congested
cell ρc , and position of the congestion front l. The model reads
1
ρ̇f = [ϕin − ϕ(ρf )] L−l






ρ̇c = [ϕ(ρc ) − ϕout ] 1l (27)



 ˙ ϕ(ρf )−ϕ(ρc )

l= ρc −ρf
where the interface flows ϕ(ρf ) and ϕ(ρc ), which correspond to the demand of
390 the free cell and the supply of the congested cell respectively, can be expressed
as
ϕ(ρf ) =ρf vmax
(28)
ϕ(ρc ) =w(ρM − ρc )
ϕin and ϕout are the inflow and outflow at the boundaries of the section of
length L. They are defined as
ϕin = min {Din , Sf }
(29)
ϕout = min {Dc , Sout }
where Din and Sout are respectively the input demand and the output supply.
395 Dc and Sf are
Dc = min {ρc vmax , vmax ρcr (vmax )}
(30)
Sf = min w(ρM − ρf ), vmax ρcr (vmax )

where ρcr (vmax ) is the critical density relative to vmax . It is defined as

wρM
ρcr (vmax ) = (31)
vmax + w
De Nunzio et al. (2014) propose to adapt the VLM to the urban environment
by considering a binary variable ζ multiplying the boundary flows in (29) to
model the behavior of traffic lights, as

1 , if the traffic light is green
ζ= (32)
0 , else
15
400 • Link transmission model
Yperman (2007) proposes the link transmission model (LTM), which is a
discrete version of the LWR model based on the triangular fundamental diagram,
with only one cell per road. Therefore, computation times are reduced.
tot
The LTM introduces the notion of cumulative vehicle counts. Nup (kδ) and
tot
405 Ndown (kδ) are respectively the cumulative entering and exiting vehicle count of
a given link at kδ, based on given split ratios at intersections.
The maximum number of vehicles that can be sent by this link to the next
one during time interval [kδ, (k + 1)δ] is

tot L tot
Sboundary (k) = Nup (k + 1)δ − − Ndown (kδ) (33)
vmax
The maximum number of vehicles that can leave the considered link during
410 the time interval [kδ, (k + 1)δ] is
Slink (k) = ρM Lδ (34)
The number of vehicles sent by the link to the next one is then simply
S(k) = min {Sboundary (k), Slink (k)} (35)
In the same way, the number of vehicles R(k) received by the link is expressed
as 

 R(k) = min {Rboundary (k), Rlink (k)}



L
tot

Rboundary (k) = Ndown (k + 1)δ − w + ρM L − Nup
tot
(kδ) (36)




Rlink (k) = ρM Lδ

• Queue models
415 Queue models are interested in the length of the queues at the end of each
link i. An example is the Berg-Lin-Xi (BLX) model, presented by Van den Berg
et al. (2007) and Lin and Xi (2008). Lin et al. (2012) propose an extension of
the BLX model. Like the LTM, the BLX model considers flows between the
links.
420 The queue on link i is composed of Niq vehicles. When the traffic light is
green, the number of vehicles entering cell i from the upstream cell during time
interval [kδ, (k + 1)δ] is
q
(k) + δϕi−1 (k), Si (k), δϕM

δϕi (k) = max 0, min Ni−1 (37)
where Si (k) denotes the available storage of link i at time step k, expressed in
number of vehicles.
425 The queue length and the available storage can be expressed as
 q q
 Ni (k + 1) = Ni (k) + δ(ϕi (k) − ϕi+1 (k))
(38)
Si (k + 1) = Si (k) + δ(ϕi+1 (k) − ϕi (k))

16
The total number of vehicles in link i at time step k can be determined as
Ni (k) = ρM L − Si (k) (39)
• Summary
The LWR model and its discrete variations, presented above, are simple first
order mathematical representations of the traffic inspired by fluid mechanics.
430 They are based on fundamental diagrams that associate ϕ to ρ. These models
are able to capture realistic traffic phenomena such as shock waves, physical
queues and queue spillbacks (Garavello et al., 2016).
However, first-order models based on the fundamental diagram are not suf-
ficient to capture unstable traffic variations caused by the inertia of vehicles
435 because they assume that v is always in equilibrium. Consequently, they have
limitations in capturing complex traffic phenomena such as stop-and-go waves,
capacity drops and phantom jams (formation of clusters of cars with high densi-
ties due to the driving style of road users Kerner and Konhäuser (1993)). These
must be taken into account in order to best estimate emissions and energy con-
440 sumption.
– Second order models
Second order models have been developed in order to capture more realistic
traffic behavior in congested areas. They still consider the equation for the
conservation of vehicles presented in (23) and use the fundamental diagram to
445 determine the steady state of the system, but they have an additional equation
for the conservation of momentum.
• Payne-Whitham model
An example of a well known second order model is proposed in Payne (1971).
The model has the following form

∂t ρ + ∂x (ρv) = 0
(40)
∂t v + v∂x v + ρ1 ∂x (p(ρ)) = τ1 (ve (ρ) − v)
450 where ve (ρ) is the equilibrium speed given by the fundamental diagram, and
p(ρ) is analogous to the pressure in the fluid dynamics equations and depends
on the density (Piccoli and Tosin, 2009).
The anticipation term ρ1 ∂x (p(ρ)) models the reaction of vehicles, i.e. accel-
eration or deceleration, to the variations of ρ. The relaxation term τ1 (ve (ρ) − v)
455 models the tendency of vehicles to travel from v towards ve (ρ) within a time
τ > 0 that represents the time needed by the vehicles to adjust their actual
speed to ve (ρ).
The second equation of (40) is the acceleration equation. Whitham (1974)
proposes to simplify the model by considering p(ρ) as a constant. Other expres-
460 sions for this term exist, they are presented in Garavello et al. (2016), as well
as the modeling of an additional viscous term in this equation.
17
• Aw-Rascle-Zhang model
Daganzo (1995) highlights some limitations of the Payne-Whitham model
presented above. In particular, the model allows the vehicles to travel with
465 negative speed.
To tackle this problem, Aw and Rascle (2000) and Zhang (2002) propose the
following model
∂t ρ + ∂x (ρv) = 0
(41)
∂t (v + p(ρ)) + v∂x (v + p(ρ)) = 0
where the pressure term p may be defined as
p(ρ) = ργ , γ>0 (42)
• METANET model
470 Messmer and Papageorgiou (1990) present the METANET model which is
a discrete version of the Payne-Whitham model presented in (40). It reads
 δ
 ρi (k + 1) = ρi (k) + δL (ϕi (k) − ϕi+1 (k))





vi (k + 1) = vi (k) + γδ1 [ve (ρi (k)) − vi (k)] (43)




+ δδL vi (k)[vi−1 (k) − vi (k)] − γ2γδ[ρ i+1 (k)−ρi (k)]


1 δL [ρi (k)+γ3 ]
and the authors propose the following fundamental diagram to define the equi-
librium speed ve (ρ)
γ
1 ρi (k) 4

ve (ρi ) = vmax exp − (44)
γ4 ρcr
where γ1 to γ4 are model coefficients.
475 METANET was originally introduced to capture traffic phenomena on high-
ways. The proceeding of flows between the segments is fully presented in Mess-
mer and Papageorgiou (1990).
– Phase transition and higher order models
Second order models generally have higher computation times. Phase tran-
480 sition models are a good alternative to the extent that they behave like the
classic LWR model when the traffic is free and like a second-order model when
the traffic is congested. This allows to capture complex traffic phenomena while
keeping reasonable computation times for free-flow traffic.
Colombo (2002) proposes the following phase transition model
485 • For free flow traffic, the author considers the LWR model, presented in
(25), with the Greenshields fundamental diagram (cf. Table 3, with p = 1).
• For congested traffic, v cannot be considered as a function only of the
density anymore. In this case, the density-flow points are scattered in a
two-dimensional region.
18
490 The associated hybrid fundamental diagram is shown in Fig. 4. The following
second-order model is then introduced for high densities

∂t ρ + ∂x (ρv) = 0
(45)
∂t q + ∂x ((q − Q)v) = 0
where q is the momentum, Q is a parameter of the road considered, and v is

expressed as
ρ q
v = 1− M (46)
ρ ρ
Figure 4: Fundamental diagram of the phase transition model, representing the free region
Ωf and the congested region Ωc (Colombo, 2002) [Copyright c 2002 Society for Industrial
and Applied Mathematics. Reprinted with permission. All rights reserved].
Finally, higher-order models exist but they are less appropriate for emissions
495 and energy consumption estimation as their computation times are higher. An
example of third order models, where the additional equation is for the variance
of the speed, can be found in Helbing (1995). This method is useful to describe
the increase of the speed variance just before traffic jams occur.
– Network-wide extension
500 The traffic models presented above can be extended across a network. The
junctions represent a very important part of the extended model. Basically, each
junction can be reduced to a combination of simple merge and diverge junctions
(Garavello et al., 2016). A complete overview of macroscopic node models can
be found in Tampère et al. (2011). The authors present macroscopic node model
505 instances both for signalized and unsignalized intersections.
In the case of the extended CTM, a fundamental diagram is associated with
each link, each link being partitioned into uniform cells. An urban version
of the CTM is proposed in Xie et al. (2013). The authors use turning ratios
assigned to intersections and distinguish two possibilities. First, a cell preceding
19
510 an intersection can be composed of one traffic light. Such cells have one unique
queue, and all the vehicles merge into it. Second, the cell can be divided into
sub cells so that each direction has its own traffic light.
Similarly, the LTM can be extended considering the flows sent and received
by links (Garavello et al., 2016). Regarding queue models, Lin et al. (2012)
515 consider the case of links with multiple junctions (connected to several upstream
and downstream links) and, for control purposes, present the S model, which is
basically a simplification of the BLX model, with a time interval equal to the
traffic-light cycle.
The network-wide extension approach is similar for second order traffic mod-
520 els. For example, Garavello et al. (2016) present the extension of the Aw-Rascle-
Zhang model on a network scale. A more detailed description of this model at
junctions can be found in Herty and Rascle (2006).
For control purposes, De Nunzio et al. (2014) suggest to simplify the VLM
by assuming an average continuous flow through the traffic lights by replacing
T
525 the binary variable ζ with Tgreen cycle
, where Tgreen and Tcycle denote respectively
the green phase time and the cycle time of the traffic light. This method is
inspired by store-and-forward models, originally suggested by Gazis and Potts
(1963). It allows to describe the urban traffic without using binary variables.
Hence, polynomial complexity control methods can be applied to the system,
530 which allows for consideration of large-scale networks. However, due to this sim-
plification, the effect of offsets between traffic lights of successive intersections is
not depicted. Moreover, the oscillations of the system (stop-and-go waves, prop-
agation waves, etc.) are not represented, which is a crucial point for emissions
and energy consumption estimation (Hall, 2012; Aboudolas et al., 2009).
535 3.2. Emission and energy consumption meta-models

In Section 3.1, we reviewed some methods to determine the traffic kinematics,
either by measuring the static average speed (cf. Section 3.1.1), or by using
dynamic fluid-based traffic models (cf. Section 3.1.2). In this section, we present
the meta-models used to calculate emissions and energy consumption from the
540 traffic dynamics, considering either approach.
3.2.1. Meta-model associated with static average speed-based approaches

Emissions and energy consumption can be calculated by considering an av-
erage speed-based approach. This is done by a meta-model whose general pro-
cedure is illustrated in Fig. 5, and operation steps are presented below.
545 1. The average speed v̄ and the number of vehicles N are measured, or esti-
mated.
2. The emission or energy consumption rate of a single vehicle Jy is calculated
from v̄ using a microscopic emission and energy consumption model (cf.
Section 2).
550 3. Jy (v̄) is then multiplied by N to approximate the total emission or energy
consumption rate.
20
v̄ Microscopic Jy (v̄) Macroscopic
Measurements emission/energy model emissions/energy
Figure 5: Structure of the emission and energy consumption meta-model associated with static
average speed-based approaches.
Note that this procedure can be conducted at different scales. The average
speed v̄ and the number of vehicles associated N can refer to a single link of
a network, if the data are available, or to a larger spatial area. Moreover, the
555 duration between two successive measurements usually depend on the measuring
devices. These issues are addressed in Section 3.3.
This meta-model can be associated either with a data-based or a physical
microscopic emission and energy consumption model. These approaches are
detailed below. Note that they involve measuring, or estimating, the number of
560 vehicles on the roads under consideration.
• Data-based model
Some authors propose to associate the meta-model with a data-based mi-
croscopic emission and energy consumption model.
For instance, Boriboonsomsin et al. (2012) propose the following regression-
565 based model in order to estimate the fuel use rate of a single vehicle

ln Jfspat 2 3 4
uel = β0 + β1 v̄ + β2 v̄ + β3 v̄ + β4 v̄ + β5 α (47)
where β0 to β5 are the regression coefficients.

Another common approach to estimate emissions and energy consumption
on a large spatial scale is to associate this meta-model with a microscopic model
based on aggregated data-driven emission or energy consumption factors Jy (v̄, θ)
570 that depend on the traffic average speed v̄ and some vehicle parameters θ.
Let Ω be the set of possible parameters sets. Aggregated factors are usu-
ally simply the mean values of experimental measurements and are typically
expressed in mass of pollutant emitted (or mass of fuel consumed) per vehicle
and per unit distance traveled. Hence, the total emission or energy consump-
575 tion rate (per distance traveled), i.e. the output of the meta-model, of a link i
containing N (i, θ) vehicles with the set of parameters θ is given by
X
Jyi = N (l, θ)Jy (v̄, θ) (48)
θ∈Ω
In practice, detailed information on the fleet composition is not available.

Hence, a reference set can be considered, i.e. all the vehicles have the same
parameters θ̄, and the emission or energy consumption rate on link i simply
580 becomes:
Jyi = N (i, θ̄)Jy (v̄, θ̄) (49)
21
The COPERT (COmputer Programme to calculate Emissions from Road
Transport) model (Ntziachristos et al., 2009) developed by the European En-
vironment Agency is based on this method. Several vehicle parameters are
included in θ: the vehicle type (passenger car, light commercial vehicle, heavy
585 duty vehicle, L-category vehicle), the fuel type, the engine displacement and its
registration date. The sets of parameters of all the vehicles constitute the vehi-
cle fleet composition. An example of emission factors obtained with COPERT
for different types of vehicle according to the speed is given in Fig. 6.
Figure 6: Fuel consumption factors of different gasoline passenger cars calculated with COP-
ERT (Sobrino et al., 2014) [Published with permission of Networks and Spatial Economics].
Hausberger (2009) proposes the HBEFA (HandBook Emission FActors for

590 road transport) model, which is more precise. This method additionally con-
siders the driving conditions (highways, urban roads, stop-and-go traffic) and
the volume-to-capacity ratio (number of vehicles divided by the capacity of the
link), which is a dynamic variable, to determine Jy (v̄, θ).
The accuracy can also be improved by multiplying the emission and energy
595 consumption factor Jy (v̄, θ) by a congestion correction factor, as does the TEE
(Traffic Energy and Emissions) model. The objective is to represent the effect
of congestion on emissions and energy consumption. The congestion correction
factor depends on the average speed, the traffic-light timing, the link length, and
the traffic density (Negrenti, 1999). These variables and parameters are used to
600 estimate the time spent in each traffic situation (cruising, acceleration, decel-
eration and idling) and thus reflect the speed variability along the considered
road link. The corresponding speed profile can then be reconstructed.
One limitation of these aggregated factors models is that the emission and
energy consumption factors are not fundamental, as they depend on the driving
22
605 cycle used during the measurements.
• Physical model
It is also possible to use this meta-model by associating it with a physical
emission and energy consumption model.
For example, Jurik et al. (2014) propose to use the following microscopic
610 physical model to estimate the energy consumption of a vehicle on link i
r
i E (i) + (ν − 1)E p (i) , if E p (i) ≤ 0
Ey (i) = Jy L(i) = (50)
E r (i) , if E p (i) > 0
where L(i) is the length of the link i, and ν ∈ [0, 1] is the downhill potential
energy recuperation coefficient. The resistance and the potential energies are
respectively given by
d
E r (i) = ACd v̄ 2 L(i) + M gCr L(i) cos α
2 (51)
E p (i) = M gL(i) sin α
To model more precisely the speed change at an intersection, De Nunzio

615 et al. (2017) introduce a transition speed at the interface between two links of
respective average speeds v̄bef ore and v̄af ter defined as
v̄bef ore + v̄af ter
vtransition = β (52)
2
where β ∈ [0, 1] is a parameter depending on the type of interface (e.g. stop
sign, traffic light, turning movement, etc.). This transition speed can be intro-
duced to any model similar to the one presented in (50) to (51) to better model
620 intersections.
3.2.2. Meta-model associated with dynamic fluid-based models

Emissions and energy consumption can be calculated by considering the
fluid-based models dynamics. This is done by another meta-model whose general
procedure is illustrated in Fig. 7, and operation steps are presented below.
625 1. First, a dynamic fluid-based traffic model is chosen (cf. Section 3.1.2). It
provides the traffic variables, i.e. ρ(x, t), v(x, t), ϕ(x, t).
2. Then, these variables are processed by an interface to generate groups of
vehicles g(x, t) sharing the same speed and acceleration. The interface
calculates the speed, acceleration and number of vehicles of each group.
630 They are respectively denoted v(g(x, t)), a(g(x, t)) and N (g(x, t)).
3. A microscopic emission and energy consumption model is chosen (cf. Sec-
tion 2). It provides the emission or energy consumption rate Jy (v, a) of
a vehicle of group g(x, t) using the outputs v(g(x, t)) and a(g(x, t)) of the
interface.
23
ρ(x, t)
v(x, t) v(g(x, t))
ϕ(x, t) a(g(x, t)) Microscopic
Macroscopic
Interface emission/energy model
traffic model
N (g(x, t)) Jy (v, a)
Macroscopic emissions/energy
Figure 7: Structure of the emission and energy consumption meta-model associated with
dynamic fluid-based traffic models.
635 4. The emission or energy consumption rate Jy (v, a) of a vehicle of group

g(x, t) is multiplied by the number of vehicles in the corresponding group
N (g(x, t)) to provide the total emission or energy consumption rate of
group g(x, t).
The procedure presented above is repeated as many times as there are
640 groups. It is important to note that this generation of groups with homoge-
neous operation variables relies on the spatial and temporal discretizations of
the traffic model. This issue is addressed in Section 3.3.
The procedure of the interface presented above is detailed in the following
(Zegeye et al., 2013).
645 To compute emissions and energy consumption, the employed traffic models
are often discrete both in time and in space. Hence, two acceleration components
have to be considered: the temporal and the spatial accelerations:
- The temporal acceleration describes the change in speed of vehicles within
a cell from one time step to the next. It only applies to the vehicles that
650 remain in the cell. It is expressed as
vi (k + 1) − vi (k)
atemp
i (k) = (53)
δ
The number of vehicles subject to this acceleration, i.e. that stay within
the cell i from time step k to time step k + 1 is equal to
Nitemp (k) = δL ρi (k) − ϕout
i (k)δ (54)
where ϕout
i (k) is the outflow of cell i during time interval [kδ, (k + 1)δ].
The first term represents the number of vehicles initially in cell i at time
655 step k.
These vehicles constitute a group g(x, t) mentioned above. There are as
many groups of this type as there are cells in the spatial discretization of
the traffic model.
24
- The spatial acceleration describes the change in speed of vehicles moving
660 from cell i to cell j. It is defined as
vj (k + 1) − vi (k)
aspat
i,j (k) = (55)
δ
The number of vehicles subject to this acceleration, i.e. that move from
the cell i to cell j during time interval [kδ, (k + 1)δ] is
spat
Ni,j (k) = ϕi,j (k)δ (56)
where ϕi,j (k) is the flow of vehicles moving from cell i to cell j.
These vehicles constitute a group g(x, t) mentioned above. There are
665 as many groups of this type as there are cells interfaces in the spatial
discretization of the traffic model.
Ultimately, a generic formulation to calculate the emissions and energy con-
sumption rate in a network made of n cells is
Pn
Jynetwork (k) = Jy (atemp
i (k), vi (k))Nitemp (k)+
Pi=1
n P n spat spat (57)
i=1 j=1 αi,j Jy (ai,j (k), vi (k))Ni,j (k)
where αi,j is a binary variable equal to 1 if cells i and j are connected in the sense
670 that vehicles can move directly from cell i to cell j, αi,j equals zero otherwise.
The first term of (57) refers to the emissions and energy consumption of vehicles
staying in the same cell from time step k to k + 1, and the second term refers
to those of vehicles moving from one cell to another.
To estimate emissions and energy consumption more precisely, this calcula-
675 tion can be done by differentiating classes of vehicles. In that case, the function
Jy can consider the real parameters of the vehicles instead of average values.
This meta-model can be associated either with a data-based or a physi-
cal microscopic emission and energy consumption model. Some examples are
presented below. Naturally, the meta-model procedure is generic and can be
680 adopted to other models.
Some authors propose to use this meta-model by associating it with a data-
based emission and energy consumption model. For example, Zegeye et al.
(2013) propose to integrate the macroscopic traffic second order model METANET
685 with the microscopic data-driven emission and fuel consumption model VT-
micro. The resulting meta-model, called VT-macro, is mainly suitable for mod-
eling emissions and energy consumption on highways.
Similarly, Lin et al. (2013) suggest to associate the traffic first order S model
with VT-micro in an urban network. The authors present a set of possible
690 behaviors for the vehicles (e.g. free, idling, accelerating, decelerating, start-
and-stop behavior). Another use of the meta-model in an urban environment
can be found in Jamshidnejad et al. (2017), in which the authors propose the
same models association.
25
• Physical model
695 It is also possible to use this meta-model by associating it with a physical
emission and energy consumption model. For example, De Nunzio et al. (2014)
develop a method based on the VLM and a physical approach to determine
energy consumption. This model considers only the spatial component of accel-
eration. In each cell, energy consumption is determined from the cell average
700 speed (either free or congested), considering zero temporal acceleration. At the
interface of the cells, the energy is calculated based on the following spatial
acceleration
vj − v i
aspat
i,j = min amax , (58)
δ
where the maximum acceleration amax is a model parameter. Note that time
does not appear in the formulation of De Nunzio et al. (2014) because the
705 analysis is performed at steady state.
3.3. Spatial and temporal discretizations

The spatial and temporal discretizations of the methods used are a crucial
point in emissions and energy consumption estimation: a compromise has to be
found between precision and computation time.
710 • Spatial discretization
Concerning the use of the average speed meta-model, measurements of aver-
age speed and number of vehicles made on a road level would naturally give bet-
ter results than measurements made on a larger spatial scale. But this depends
mainly on the devices used to monitor the traffic. Some average speed-based
715 meta-models consider a fine spatial discretization in order to be compatible with
urban networks (e.g. COPERT Street Level (Rai et al., 2017)).
When using the meta-model associated with dynamic fluid-based models,
the choice of the spatial discretization step size should be given some thought.
A balance concerning the number of cells and their length has to be found in
720 order to satisfy the desired accuracy without excessively increasing computation
times.
• Temporal discretization
Some authors have proposed methods to use the average speed meta-model
with high-frequency data inputs, i.e. average speeds and number of vehicles
725 updated at high frequency. For example, Lejri et al. (2018) propose a method
to adapt the COPERT emission and fuel consumption model to high-frequency
data inputs. This kind of approach is more precise. However, it is essential
to note that the average speed-based meta-model is static. In other words,
even with high frequency data inputs, emissions and energy consumption are
730 calculated for successive average speeds, but do not consider the acceleration
of vehicles, yet crucial to fully characterize emissions and energy consumption
(Ahn and Rakha, 2008).
26
The dynamic fluid-based meta-model considers vehicles accelerations. There-
fore, this approach is more precise, but its crucial point is the estimation of the
735 acceleration. In the formulation proposed in Section 3.2.2, the choice of δ must
lead to realistic accelerations values while respecting the Courant–Friedrichs–Lewy
(CFL) condition expressed as
2δvmax ≤ δL (59)
• Summary
In other words, adopting a dynamic fluid-based approach with very long
740 time step size and length of cells is similar to having an average speed-based
approach. The difference would be that the data are obtained by simulation
instead of being measured.
Finally, the most precise way to calculate emissions and energy consump-
tion at a large spatial scale would be to use a microscopic traffic model and
745 to associate it with a microscopic emission and energy consumption model (cf.
Section 2). In fact, this approach is the only one able to reflect differences in
microscopic drivers’ behavior (e.g. sudden deceleration, merging, lane chang-
ing). When traffic is congested, these can result in shock waves causing traffic
breakdown, that a macroscopic traffic model cannot depict (Khondaker and
750 Kattan, 2015b). However, this method is not possible at large scale because
of the enormous computation times generated by the large number of vehicles
considered. Schiper (2017) proposes a statistical approach to process this large
amount of data by introducing sampling methods. The author suggests to esti-
mate emissions and energy consumption only in some relevant locations of the
755 network, and to extend the estimations at larger scales.
4. Single vehicles control design for emission and energy consumption

reduction
In Sections 2 and 3, emission and energy consumption models have been pre-
sented for single vehicles and for traffic flows. In this section, we review some
760 control strategies for single vehicles aiming at limiting emissions and energy
consumption. They can be mostly categorized into eco-driving, i.e. computing
a vehicle speed trajectory that minimizes the emissions or energy consumption
along a given route, and eco-routing, i.e. planning a minimum energy or emis-
sions route. An excellent overview of the existing vehicle control strategies is
765 given by Guanetti et al. (2018).
4.1. Eco-driving
Eco-driving consists in computing a vehicle trajectory that minimizes the
emissions or energy consumption along a given route, under technical (speed,
acceleration and brake characteristics of the vehicle) and environment (traffic,
770 traffic signs, traffic lights, etc.) constraints (Guanetti et al., 2018).
27
We define the state vector of a vehicle at time step k as x(k) = [s(k), v(k)]T ,
where s and v respectively denote the vehicle’s position along the route and the
speed of the vehicle. Let Ftrac and Fb be respectively the traction force at the
wheels and the mechanical brake force, the objective of the eco-driving algorithm
775 is to find at each time step k the input vector u(k) = [Ftrac (k), Fb (k)]T that
minimizes the emissions or the energy consumption calculated by the function
g. The objective function g is similar to those presented in Section 2. It may
consider vehicle parameters like its mass and parameters of the environment like
the road slope, usually provided by a Geographic Information System (GIS).
780 Sciarretta et al. (2015) formulate the eco-driving optimization problem as
follows
n−1
X
minimize g(x(k), u(k)) (60)
u0 ,...,un−1
k=0
subject to

x(k + 1) = f (x(k), u(k)), 

0 ≤ s(k) ≤ sf ,



vmin (k, s(k)) ≤ v(k) ≤ vmax (k, s(k)), ∀k ∈ [0 .. n − 1]
Ftrac,min (v(k)) ≤ Ftrac (k) ≤ Ftrac,max (v(k)), (61)




Fb,min ≤ Fb (k) ≤ Fb,max ,

x(0) = x0 ,
x(n) = xf .
The state of the vehicle at time step k + 1 is given by the following function
based on the vehicle dynamics

s(k) + δv(k)
f (x(k), u(k)) = δ (62)
v(k) + M (Ftrac (k) − Fb (k) − Fres (k))
785 where Fres is the resistance force, expressed in (8). The technical limits of the
vehicle consist in bounding the input variables Ftrac and Fb as indicated in
(61). The function vmin and vmax define a convex constraint set that takes into
account the environment constraints associated with speed limits, traffic lights,
traffic signs, road curvature, etc. x0 = [s0 , v0 ]T and xf = [sf , vf ]T are the
790 initial and final constraints of the eco-driving problem.
The eco-driving problem considering Ftrac and Fb as control inputs is per-
fectly compatible with autonomous vehicles, which include the control in the
longitudinal and lateral directions, as it gives instructions to the powertrain.
However, it is expected from an eco-driving problem for human drivers to re-
795 turn an advisory speed profile the user can follow. In that case, the algorithm
may return at each time step k the speed instruction v(k + 1) calculated with
(62) instead of Ftrac (k) and Fb (k). Another solution is to directly formulate
the optimization problem considering the recommended maximal speed of the
vehicle as the control input (Ozatay et al., 2014a; Boehme et al., 2013).
800 Sciarretta et al. (2015) present several algorithms aiming at solving the eco-
driving problem given in (60) to (62). These solutions can either be offline, i.e.
28
consider all road characteristics known in advance, or online, i.e. make use of
real-time estimations on a vehicle immersed in its environment.
Many solutions can be used for offline optimization: dynamic programming
805 (Dib et al., 2012), Pontryagin’s minimum principle (Sciarretta et al., 2015) or
calculating the analytical solution (Ozatay et al., 2014b).
Online solutions allow to acquire more information in real time about the
upcoming route. For example, Hellström et al. (2009) propose a method with an
on board optimizing controller taking into account the road slope. In the case
810 of connected vehicles, one may also imagine a control design taking into account
the prediction of the upcoming traffic conditions and accordingly updating the
vmin and vmax constraints of (61). The main limitation of these online solutions
is the computation time as they are expected to be compatible with real-time
execution.
815 In an urban environment, eco-driving is complex because of the uncertainty
of traffic. In particular, it is very difficult to know the traffic-light cycles in ad-
vance as some signalized intersections have a variable phase duration depending
on the traffic level. Intelligent transportation systems and traffic infrastruc-
ture connectivity are expected to reduce this uncertainty (Dimitrakopoulos and
820 Demestichas, 2010).
If the traffic-light cycles are unknown by the eco-driving algorithm, Ozatay
et al. (2014a) propose a method that considers traffic lights as stop signs in the
optimization problem. Naturally, the driver is free not to follow the advised
velocity given by the algorithm in the case of green at a traffic light.
825 To take into account the uncertainty about traffic-light cycles, Sun et al.
(2018) consider a stochastic cycle timing that adds to the red-light duration a
random variable. To generate more realistic signal timings, Mahler and Vahidi
(2012) introduce for each intersection a time-varying probability of green based
on measured data. In the optimization process, solutions that pass through
830 time intervals with high green probability are then naturally preferred.
In the case of known and deterministic traffic-light cycles, many algorithms
can be used to solve the eco-driving problem. For example, Miyatake et al.
(2011) present a method based on dynamic programming, De Nunzio et al.
(2016) use Dijkstra’s shortest path algorithm, HomChaudhuri et al. (2017) de-
835 velop a method with model predictive control, and Seredynski et al. (2013)
implement a genetic algorithm. The principle of these algorithms is to add a
constraint on the crossing time at intersections.
To improve the safety and avoid rear-end collisions, Zhang and Cassandras
(2018) propose a control strategy for vehicles crossing an urban signal-free in-
840 tersection. The principle is to generate acceleration profiles for the vehicles in
order to cross the merging zone in a limited time while minimizing the accel-
eration. This approach is adapted for autonomous vehicles, but the authors
consider a mixed traffic in their simulation (autonomous and human-piloted
vehicles) and analyze the impact of the proportion of autonomous vehicles on
845 their acceleration. Human-piloted vehicles are subject to priority rules.
Eco-driving algorithms need information about the traffic situation in order
to be accurate. These data can be provided either by sensors or by a macroscopic
29
traffic model. Many parameters of the problem, like pedestrians or drivers
decision making, remain uncertain and unpredictable.
850 Autonomous vehicles raise the issue of their safety, but they offer prospects in
terms of energy savings as they can accurately track the instructions generated
by the eco-driving algorithm (Han et al., 2018). Moreover, if the autonomous
vehicles communicate with each other, they can reduce their energy consumption
by coordinating and forming micro-platoons along the route, even if they have
855 different origins and destinations (Lelouvier et al., 2017).
4.2. Eco-routing
Eco-routing consists in planning an emission or energy-minimal route, given
an origin and a destination. The function that attributes to each link the energy
consumption (or the emissions) of a vehicle traveling along this link is denoted
860 g.
In the case of static eco-routing algorithms, the function g depends only on
the link under consideration. In the general case, the function g depends on the
time t as the traffic situation in the network evolves over time.
Ericsson et al. (2006) present an eco-routing algorithm that classifies the
865 roads of the network into 6 groups, depending on their GPS data. Based on the
same data, a fuel consumption factor is calculated for each group. Then, the
function g assigns to each link its energy consumption, using the fuel consump-
tion factor and the length of the link. The authors introduce peak and off-peak
hours to model the evolution of the traffic during the day. Similarly, Boriboon-
870 somsin et al. (2012) propose to consider not only historical GPS data, but also
real-time vehicle velocity trajectories to estimate the energy consumption of
each link, i.e. build the function g.
Usually, eco-routing algorithms only take into account the energetic cost of
links and not the vehicle behavior at intersections. However, this aspect is cru-
875 cial in energy consumption estimation. To model the energy consumption at
intersections, De Nunzio et al. (2017) introduce a transition speed at the inter-
face between two links, given in (52). Traffic lights at intersections have also to
be considered. For example, Sun and Liu (2015) propose an eco-routing algo-
rithm based on a signalized traffic network in which the authors use a Markov
880 decision process to model the traffic.
To determine the energy-optimal route, heuristic searches can be imple-
mented (Nannicini et al., 2012). Kluge et al. (2013) propose another approach
as the authors solve a time-dependent eco-routing problem by using an extension
of Dijkstra’s algorithm.
885 In order to limit the computation time, eco-routing algorithms can consider
a constraint on the maximum travel time or distance to reduce the set of possible
solutions. Another possibility is to implement multi-objective eco-routing that
minimizes not only the energy consumption but also the travel time and distance
traveled. In this case, the solution proposed is a Pareto optimal route (Bertsekas,
890 1995; De Nunzio et al., 2017).
30
5. Traffic flow control design for emission and energy consumption
reduction
In Section 4, we presented vehicle-based control designs aiming at reducing

the emissions and energy consumption of a single vehicle. In this section, we
895 review some road-based control strategies to reduce the environmental impact
on a large spatial scale and for a large number of vehicles. These strategies
consist in regulating vehicular flow by controlling speed limits, traffic-light duty
cycles or offsets, split ratios at intersections or bifurcations, or mobile actuators
(e.g. autonomous vehicles). In the following, the control strategies are classified
900 according to the employed actuator. For each actuator, we review first the
strategies adapted to highways, and then the strategies that can be set up for
an urban environment.
The objective of the following frameworks is to determine via an optimization
method, at each control time step, the control inputs that minimize the traffic
905 emissions and energy consumption. Note that some of the papers presented
in the following do not explicitly minimize emissions or energy consumption.
Instead, they tend to mitigate congestion and eliminate shock waves through
density homogenization, vehicles interdistance equalization, etc. These methods
are likely to indirectly reduce emissions and energy consumption as they reduce
910 the number of acceleration and deceleration (Barth and Boriboonsomsin, 2008).
However, it is important to make careful analyses about the assessment of the
effect of congestion reduction on emissions and energy consumption. In fact,
their relationship depends on many factors such as the speed of the traffic (Fiori
et al., 2018).
915 5.1. Speed limits control

A first approach to regulate the flow in order to reduce the emissions and
energy consumption is to control speed limits. This corresponds to imposing
variable location-dependent speed limits across the road network.
Many works present variable speed limits adapted to highways. Some of
920 them do not aim at explicitly reducing emissions and energy consumption (e.g.
SPECIALIST method that eliminates shock waves (Hegyi et al., 2008)). An
increasingly common approach is to use reinforcement learning methods to op-
timize speed limits. Walraven et al. (2016) propose to follow this approach to
minimize the amount of time vehicles spend on the highway under consideration.
925 Some other works are explicitly oriented on emissions and energy consump-
tion reduction. Generally, they implement a multi-objective optimization that
minimizes also the travel time so that unrealistic solutions like speed limits
equal to zero are avoided. For example, Zu et al. (2018) express the energy con-
sumption minimization on highways as a convex quadratic optimization prob-
930 lem whose objective function is derived from the average speed-based COPERT
model. The density is expressed as a function of the speed, considering the
Greenshields fundamental diagram (cf. Table 3). Another approach is proposed
by Zegeye et al. (2012). The authors proposes a control design applicable to
31
highways in which the speed limits are determined by Model Predictive Con-
935 trol (MPC). Because of the non-convex nature of the objective function, Zegeye
et al. (2012) uses a multi-start local sequential quadratic programming method
to determine the control inputs.
MPC offers opportunities for traffic control as it is compatible with the
uncertainties of the traffic models, and it can also handle non-linear and non-
940 convex optimization. However, computation times have to be reduced to make
MPC tractable for real-time operation, especially when the number of control
inputs is too large. Hence, in order to use MPC for macroscopic traffic con-
trol without significantly compromising the performance, Zegeye et al. (2012)
proposes to use a parameterized MPC, more specifically called Rolling Horizon
945 Parameterized (RHP) control.
In RHP control, the control inputs are parameterized according to some
time-profiles and the optimization focuses on the parameters. The number of
parameters to optimize is smaller than the number of control inputs and the
set of possible solutions is generally smaller. This results in faster computation
950 times but also a loss of performance. For computation time issues, RHP control
is more suitable for real-time application than conventional MPC, but it still
may be too slow, depending on the considered system, the parameterization,
and the control time step.
Note that some authors propose to use approaches based on microscopic
955 traffic models to control variable speed limits on freeways. For example, Khon-
daker and Kattan (2015a) present an MPC-based approach to maximize mobil-
ity, safety and environmental benefit.
A hybrid approach proposed in Van den Berg et al. (2007) aims at controlling
speed limits for mixed urban and highway networks. The authors present an
960 MPC framework that minimizes the TTS.
In an urban environment, some works do not explicitly minimize emissions
and energy consumption. For example, Tajali and Hajbabaie (2018) present
an MPC framework aiming at harmonizing the speed within the network and
maximizing the outflows.
965 Other works explicitly aim at reducing emissions and energy consumption.
Taylor (2000) presents an approach to evaluate the impact of various speed
limits on emissions, energy consumption, and traffic congestion, without seeking
to optimize speed limits. De Nunzio et al. (2014) propose a method to find the
optimal speed limit of a road section. The traffic model considered is the VLM
970 presented in Section 3.1.2, and the same notations are used here. The control
input is vmax and the objective function is the weighted sum of the energy
consumption, the total time spent T T S, the instantaneous travel time IT T and
the total travel distance T T D. These new metrics are defined as follows
L−l l
IT T (ρ) = + (63)
vmax vc
T T D(ρ) = Tcycle ρf vmax L + [w(ρM − ρc ) − vmax ρf ]l

(64)
32
T T S(ρ) = Tcycle (ρf L + (ρc − ρf )l) (65)
A method based on shock waves theory to control speed limits in an urban
975 area has been proposed by De Nunzio and Gutman (2017) to optimize energy
consumption and T T S.
Panis et al. (2006) present a methodology to analyze the environmental im-
pact of speed limits in an urban environment. The authors use the microscopic
traffic model DRACULA and a data-based emission and energy consumption
980 model. A case study is conducted in Ghentbrugge, a neighborhood of the city
of Ghent, Belgium. Similarly, Liu and Tate (2004) propose to study the effect
of speed limits in an urban network by implementing Intelligent Speed Adapta-
tion (ISA). This system suggests, or imposes, speed limits to the driver through
in-vehicle electronic devices. Note that ISA only informs road users of the speed
985 limits, but does not calculate it independently for each vehicle. In other words,
it is just a communication device. In this study, the authors consider the speed
limits as inputs of the simulation, i.e. they can vary with locations but are fixed
over the simulation time period, and are not optimized. One may also consider
dynamic speed limits based on an optimization framework. The authors use the
990 DRACULA traffic model. One of the main limitations of this kind of approach
based on microscopic traffic models is that a lot of data are involved. They are
usually very difficult to obtain, and they cause long computation times.
Note that machine learning methods can be used to control variable speed
limits. For example, Zhu and Ukkusuri (2014) present a Reinforcement Learning
995 (RL) approach aiming at optimizing the total network throughput, the delay
time, and the emissions. The authors propose a case study conducted on the
Sioux Falls network.
A general overview of the theoretical background and the main strategies of
variable speed limits strategies can be found in Khondaker and Kattan (2015b).
1000 5.2. Traffic lights control

Another solution to reduce emissions and energy consumption using road-
based control is traffic signal timing optimization. On freeways, this control
strategy, known as ramp metering, can be applied on on-ramps, and it consists
in regulating the traffic flow entering the highway.
1005 Many ramp metering strategies do not explicitly optimize emissions and
energy consumption, but they aim at reaching a desired density. That is the
case of ALINEA method, presented in Papageorgiou et al. (1991), which uses a
feedback law and the traffic density measured downstream from the merge area.
Similarly, Pisarski and Canudas-de-Wit (2016) present an approach to balance
1010 the vehicle density on the freeway by formulating the optimization problem as
a non-cooperative Nash game.
Some authors express the ramp metering control approach as an optimiza-
tion problem aiming at directly reducing emissions and energy consumption. For
example, Csikós et al. (2011) present a multi-objective optimization based on
1015 a constrained LQ (Linear-Quadratic) control, minimizing both T T S and traffic
33
emissions on freeways. Pasquale et al. (2015) formulate the ramp metering con-
trol problem as a multi-objective nonlinear constrained optimization problem
considering the same objective function. These metrics are calculated consid-
ering both the traffic in the on-ramp and in the mainstream. The emissions
1020 are calculated using an average speed-based model based on COPERT. The
nonlinear optimization problem is solved with a specific version of the feasible
direction algorithm: the derivative backpropagation method RPROP. A specific
feature of this work is that the authors consider two classes of vehicles (cars and
trucks) individually controlled as determined by the optimization process.
1025 In an urban environment, many traffic signal timing optimization strategies
have been developed to control traffic-light cycles. Most of them do not explic-
itly optimize emissions and energy consumption. Instead, they minimize the
congestion by improving the throughput and reducing the delay. Some of these
strategies are: SCOOT (Split, Cycle and Offset Optimisation Technique) (Hunt
1030 et al., 1981), SCATS (Sydney Coordinated Adaptive Traffic System) (Lowrie
et al., 1982), RHODES (Real-time Hierarchical Optimized Distributed Effec-
tive System) (Mirchandani and Head, 2001), TUC (Traffic-responsive Urban
Control) (Dinopoulou et al., 2006), max-pressure (Varaiya, 2013).
Grandinetti et al. (2018) formulate the signal timing control problem as
1035 a CTM-based real-time convex optimization whose objective function is the
weighted sum of T T D, the density balancing and a regularization term that
penalizes abrupt changes in the control dynamics. The density balancing term
aims at homogenizing the density over the network. The algorithm is split into
subproblems whose sizes are independent of the network size, thus allowing for
1040 scalability.
Some authors explicitly consider emissions and energy consumption reduc-
tion. For example, Han et al. (2016) express the signal timing optimization in
an urban environment as an LTM-based Mixed Integer Linear Program (MILP)
optimizing both the delays and the emissions. Emissions are calculated as a
1045 function of the density of the links, by calculating the spatial and the tempo-
ral accelerations defined in Section 3.2.2. Similarly, Osorio and Nanduri (2015)
propose a meta-model that considers the simulations of T T S and fuel consump-
tion as well as their analytical approximations to solve the urban signal timing
optimization using a simulation-based optimization algorithm.
1050 MPC can be implemented in a traffic light control framework. For example,
Lin et al. (2013) present a method adapted for urban traffic networks based on
MPC. The authors consider a dynamic fluid-based meta-model associating the
S model and VT-micro to characterize emissions (cf. Section 3.2.2). The ap-
proach aims at reducing both the congestion and the emissions as the objective
1055 function considers the weighted sum of T T S and total emissions. Jamshidnejad
et al. (2018) propose a similar approach based on a gradient-based optimiza-
tion approach. The authors consider an extension of the S traffic model. The
objective function considers the weighted sum of T T S, total emissions, and the
absolute difference of two temporally successive control inputs, in order to avoid
1060 abrupt variations.
A more precise approach can be found in Stevanovic et al. (2009). The
34
authors propose to simulate the traffic dynamics through a microscopic traffic
model, namely VISSIM, and to calculate the emissions by using the CMEM
emission and energy consumption model (cf. Section 2.2). A signal timings op-
1065 timization is then conducted using VISGAOST, an optimization program based
on the stochastic nature of genetic algorithms. Although the authors propose
a case study on a road network composed of two suburban arteries, an online
optimization based on this method is not possible because of lengthy calculation
times. But such methods can be implemented for an offline optimization.
1070 RL methods can also be implemented to control traffic lights in an urban
network. For example, Khamis and Gomaa (2012) present a framework that
considers the microscopic dynamics of vehicles. The authors propose to approx-
imate the energy consumption metric by the average number of vehicles stops,
assuming that this performance index can be directly related to ecological issues.
1075 5.3. Coordinated speed limits and traffic lights control

To improve the results of road-based control, it is possible to coordinate
actuators such as speed limits and signal timing control.
For freeways control, Hegyi et al. (2005) propose a method to optimize T T S,
without considering emissions and energy consumption. The authors develop a
1080 MPC framework, in which the control inputs are speed limits and ramp meter-
ing.
Other authors have used coordinated speed limits and signal timing control
to reduce emissions and energy consumption. For example, Zegeye (2011) opti-
mizes T T S, the fuel consumption, and N Ox emissions via MPC. A very similar
1085 approach is presented in Liu et al. (2017). The authors also use MPC to control
both ramp metering and speed limits on a highway section, and the objective
function is the weighted sum of T T S and the total emissions. A specific feature
of this work is that multiple classes of vehicles are considered.
A problem to study in an urban environment is bandwidth maximization
1090 along an artery. Assuming that all the traffic lights have a common cycle, the
problem of bandwidth maximization consist in maximizing the vehicle through-
put along the artery under study, by traffic lights offset control. Usually, the
actuators are only the traffic lights offset, like presented in Mehr et al. (2018) in
which the authors express a nonlinear optimization problem and convert it to a
1095 MILP. The bandwidth maximization problem optimizes the flow of vehicles but
does not explicitly reduce the emissions and energy consumption. Therefore,
De Nunzio et al. (2015) propose a formulation as an optimization problem in
which the objective function contains also terms approximating T T S and energy
consumption. In this work, the authors use coordinated actuators, namely speed
1100 limits and signal timing controls. None of the bandwidth maximization strate-
gies presented is based on a traffic model. Hence, they work best in steady-state
under-saturated traffic conditions.
5.4. Dynamic routing

Another solution to reduce emissions and energy consumption is to use dy-
1105 namic routing. This method consists in redistributing the traffic demand over
35
the network in a more efficient way by controlling the split ratios. In practice,
the controller predicts the optimal routes for the main traffic flow directions,
and the associated recommendations are communicated to the road users by
the mean of in-vehicle devices, radio, or variable message signs (Treiber and
1110 Kesting, 2013a).
In the literature, the control objective of dynamic routing problems is usu-
ally to reach system-optimum or user-equilibrium. The system-optimum corre-
sponds to the minimum T T S and the user-equilibrium is characterized by a den-
sity distribution for which all used routes between the same origin-destination
1115 pair have the same travel time (Xu et al., 2011).
Dynamic routing could also be used to directly reduce emissions and en-
ergy consumption. For example, Luo et al. (2016) propose a real-time en-route
diversion control strategy that minimizes T T S, the total emissions and fuel
consumption. The route recommendation provided by variable message signs
1120 is considered as the control variable. The split ratios are calculated from the
route recommendation considering a drivers’ compliance rate which is supposed
to be known. The route diversion control uses MPC based on a parallel Tabu
Search algorithm.
Emission pricing can also be used as a dynamic routing method aiming at
1125 influencing route selection in order to reduce emissions and energy consumption.
This method can be static or dynamic. Dynamic road pricing studies based on
emissions and energy consumption are reviewed in Wang et al. (2018).
5.5. Mobile actuators

Most of the strategies presented above are motionless in the sense that traffic
1130 lights, ramp metering, message signs and speed limit signs exert commands at a
fixed location. A new approach is to consider mobile actuators, namely vehicles
that could be controlled to have an impact on the surrounding traffic. Typically,
this corresponds to the injection of some autonomous vehicles in the traffic flow
with the objective of stabilizing it.
1135 Stern et al. (2019) present how this method can reduce the emissions of the
whole traffic by dampening stop-and-go waves. To validate this approach, the
authors present the results of field experiments in which vehicle velocity and
acceleration data are collected. These experiments use a single autonomous
vehicle to dampen traffic waves on a ring road with 20 other human-piloted
1140 vehicles. The results are coherent with the simulations of Wu et al. (2018). Yang
and Jin (2014) present a similar control based on inter-vehicle communication.
Autonomous vehicles present opportunities in terms of traffic stabilization,
emissions, and energy consumption. They also induce a smoother driving and
fewer braking events. But the results presented in Stern et al. (2019) hold for
1145 situations with traffic waves only.
A country-level evaluation of the impact of autonomous vehicles on the en-
vironment can be found in Liu et al. (2019). The authors consider different
scenarios regarding the autonomous vehicle penetration rate by 2050.
36
6. Conclusion and outlook
1150 The current situation regarding pollutant emissions and energy consumption
of road transportation is alarming both for environmental and health reasons.
Ecological traffic management appears to be a promising lever in the long-term
to reduce the environmental impact of transportation.
This paper surveys the existing emission and energy consumption models,
1155 as well as the traffic control strategies to reduce them, either by considering
vehicles independently, or by considering traffic flows. The main advantages
and drawbacks of the different approaches are highlighted.
The first step to estimate emissions and energy consumption is to measure,
or simulate, the kinematics of vehicles, that can be either static or dynamic.
1160 Traffic models can be implemented on a microscopic or macroscopic scale. The
complexity of large scale road networks is essentially due to to the processing of
junctions, and the choice of temporal and spatial discretizations, which represent
a crucial point.
From the traffic kinematics, emissions and energy consumption can be esti-
1165 mated using either data or physics-based approaches. Thus, many associations
of models are possible. For complexity reasons, some are more suitable than
others. In order to go large scale, the objective is to find a balance between
accuracy and computation time, which depends mainly on the use of the frame-
work (e.g. compatibility with control methods). For example, a question is
1170 whether the additional complexity introduced by a second-order traffic model
significantly improves the accuracy of a first order model in depicting the traffic
behaviors that impact energy efficiency. Similarly, a microscopic approach to
describe large-scale emissions and energy consumption would provide the best
estimations but it would involve a lot of data that can be difficult to obtain
1175 and process, the need to precisely calibrate the model, and a sharp increase in
computation times. However, this approach can be useful for offline validation
purposes.
Traffic management can be carried out by controlling a single vehicle to
reduce its emissions and energy consumption, or by acting on a large spatial
1180 scale with actuators such as speed limits, traffic lights, dynamic routing or
autonomous vehicles. Usually, a multi-objective optimization is considered to
control the traffic with ecological concerns in order to ensure realistic solutions.
Many articles propose traffic control designs that do not aim at reducing
emissions and energy consumption but more classic metrics such as the distance
1185 traveled, the delays or the total time spent in the networks. These methods can
be adapted to multi-objective control problems considering ecological issues,
which offers promising opportunities in this research field. Also, note that some
control designs aiming at improving traffic fluidity can have a positive impact
on environmental metrics as they reduce the number of stops and accelerations.
1190 Some clear trends can be identified in the ecological approach of traffic con-
trol. For example, autonomous vehicles are considered the next major techno-
logical advance in the transport sector. Not only do they have an important
role to play in road safety, but they can also reduce the impact of transport
37
on the environment by reducing vehicle ownership and improving energy con-
1195 sumption rate (Liu et al., 2019). However, at system-wide level, the effect of
autonomous vehicles on travel demand and energy efficiency is very uncertain
and might increase the total fuel consumption (Brown et al., 2014; U.S. De-
partment of Energy, 2018). Autonomous vehicles are a lever able to influence
and regulate the surrounding traffic. Analyses of the best penetration rate of
1200 connected and automated vehicles in freeway traffic to improve global energy
performance are recently appearing (Rios-Torres and Malikopoulos, 2018), and
further exploration of the effects in an urban environment should be conducted.
Connected vehicles able to communicate with the infrastructure are also
expected to become more numerous, which would considerably increase the
1205 available data. Moreover, computing capabilities have recently been greatly im-
proved. These aspects are expected to improve the efficiency of control strategies
and make microscopic approaches a more interesting option.
Machine learning methods are also a major trend both for estimation and
reduction purposes of emissions and energy consumption. Neural networks are
1210 becoming more and more precise to estimate emissions and energy consumption
from vehicle operating variables. They are a good alternative to physics-based
approaches because of the high non-linearity of emissions and energy consump-
tion. To control the traffic, by any actuator, RL approaches are an important
trend.
1215 MPC approaches seem to be very popular among traffic control methods
as it is compatible with the uncertainties of the traffic models, and it can also
handle non-linear and non-convex optimization.
Some gaps can be identified concerning the ecological approach of traffic
control. One of them is that obtaining large-scale data is difficult because most
1220 vehicles are still not fully connected. The issue of missing data imputation has
been addressed by some authors (Qu et al., 2009).
Moreover, experiments in traffic control are very long and expensive to put
in place. Hence, most approaches are not validated by real experimentation.
However, more and more cities are taking action to reduce pollutant emissions.
1225 For example, many cities are generalizing the speed limit to 30 km/h in most
streets (Bordarie, 2017). Similarly, old diesel vehicles are being banned from
many large cities, especially in Germany (Möhner, 2018). These strong mea-
sures could be associated with a dynamic control of traffic aiming at explicitly
reducing emissions and energy consumption.
1230 Another identified gap is that models intended for large-scale control pur-
poses are limited by computation time. A major issue is their level of detail
(e.g. approximation of the acceleration of macroscopic traffic models, process-
ing of junctions in road networks). This determines the compromise between
accuracy and computation time, which inevitably leads to approximations. In a
1235 hypothetical future in which many vehicles would be connected or autonomous,
the question of data processing from a computational point of view for control
purposes would also certainly arise.
Concerning control strategies, a moot point is to find the metric to optimize
along with the ecological issues. Depending on the objective, many approaches
38
1240 are possible (e.g. minimize the travel duration or distance, homogenize the
density or the speed).
An interesting aspect to study would be the impact of traffic congestion
on emissions and energy consumption. For example, it could be interesting to
analyze in detail the most impacting traffic phenomena on emissions and energy
1245 consumption.
The best models for emission and energy consumption reduction for large-
scale road networks are probably yet to be found. To control the traffic with
ecological concerns, one may explore the use of new actuators, or coordinate
them at a large spatial scale.
1250 Acknowledgements
This project has received funding from the European Research Council
(ERC) under the European Union’s Horizon 2020 research and innovation pro-
gramme (grant agreement 694209).
39
References
1255 Aboudolas, K., Papageorgiou, M., and Kosmatopoulos, E. (2009). Store-and-

forward based methods for the signal control problem in large-scale congested
urban road networks. Transportation Research Part C: Emerging Technolo-
gies, 17(2):163 – 174.
Ahn, K. (1998). Microscopic fuel consumption and emission modeling. Master’s
1260 thesis, Virginia Tech.
Ahn, K. and Rakha, H. (2008). The effects of route choice decisions on ve-
hicle energy consumption and emissions. Transportation Research Part D:
Transport and Environment, 13(3):151 – 167.
Ahn, K., Rakha, H., and Trani, A. (2004). Microframework for modeling of
1265 high-emitting vehicles. Transportation Research Record, 1880:39–49.
Alsabaan, M., Naik, K., Khalifa, T., and Nayak, A. (2012). Applying vehic-
ular networks for reduced vehicle fuel consumption and co2 emissions. In
Intelligent Transportation Systems. IntechOpen.
An, F., Barth, M., Norbeck, J., and Ross, M. (1997). Development of com-
1270 prehensive modal emissions model: operating under hot-stabilized condi-
tions. Transportation Research Record: Journal of the Transportation Re-
search Board, (1587):52–62.
Andersen, O., Jensen, C. S., Torp, K., and Yang, B. (2013). Ecotour: Re-
ducing the environmental footprint of vehicles using eco-routes. In Mobile
1275 Data Management (MDM), 2013 IEEE 14th International Conference on,
volume 1, pages 338–340. IEEE.
Andre, M. and Pronello, C. (1997). Relative influence of acceleration and speed

on emissions under actual driving conditions. International Journal of Vehicle
Design, 18(3-4):340–353.
1280 Aw, A. and Rascle, M. (2000). Resurrection of” second order” models of traffic
flow. SIAM journal on applied mathematics, 60(3):916–938.
Barth, M., An, F., Norbeck, J., and Ross, M. (1996). Modal emissions mod-
eling: A physical approach. Transportation Research Record: Journal of the
Transportation Research Board, (1520):81–88.
1285 Barth, M. and Boriboonsomsin, K. (2008). Real-world carbon dioxide impacts
of traffic congestion. Transportation Research Record: Journal of the Trans-
portation Research Board, (2058):163–171.
Bertsekas, D. P. (1995). Dynamic programming and optimal control, volume 1.
Athena scientific Belmont, MA.
40
1290 Boehme, T., Held, F., Rollinger, C., Rabba, H., Schultalbers, M., and Lampe,
B. (2013). Application of an optimal control problem to a trip-based energy
management for electric vehicles. SAE International Journal of Alternative
Powertrains, 2:115–126.
Bordarie, J. (2017). Public policy of urban mobility: Impact of the history
1295 and practices on young drivers’ social representation of 30 km/hr. Journal of
Nonprofit & Public Sector Marketing, 29(2):211–234.
Boriboonsomsin, K., Barth, M. J., Zhu, W., and Vu, A. (2012). Eco-routing nav-
igation system based on multisource historical and real-time traffic informa-
tion. IEEE Transactions on Intelligent Transportation Systems, 13(4):1694–
1300 1704.
Brown, A., Gonder, J., and Repac, B. (2014). An analysis of possible energy
impacts of automated vehicles. In Meyer, G. and Beiker, S., editors, Road
vehicle automation, Lecture notes in mobility, pages 137–153. Springer, Cham.
Burghout, W. (2004). Hybrid microscopic-mesoscopic traffic simulation. PhD
1305 thesis, KTH.
Canudas-de-Wit, C. (2011). Best-effort highway traffic congestion control via
variable speed limits. In Decision and Control and European Control Confer-
ence (CDC-ECC), 2011 50th IEEE Conference on, pages 5959–5964. IEEE.
Canudas-de-Wit, C. and Ferrara, A. (2018). A variable-length cell transmission

1310 model for road traffic systems. Transportation Research Part C: Emerging
Technologies, 97:428 – 455.
Chapman, L. (2007). Transport and climate change: a review. Journal of
Transport Geography, 15(5):354–367.
Colombo, R. M. (2002). Hyperbolic phase transitions in traffic flow. SIAM

1315 Journal on Applied Mathematics, 63(2):708–721.
Csikós, A., Luspay, T., and Varga, I. (2011). Modeling and optimal control
of travel times and traffic emission on freeways. IFAC Proceedings Volumes,
44(1):13058 – 13063. 18th IFAC World Congress.
Daganzo, C. F. (1994). The cell transmission model: A dynamic representation

1320 of highway traffic consistent with the hydrodynamic theory. Transportation
Research Part B: Methodological, 28(4):269–287.
Daganzo, C. F. (1995). Requiem for second-order fluid approximations of traffic
flow. Transportation Research Part B: Methodological, 29(4):277–286.
De Nunzio, G., Canudas-de-Wit, C., and Moulin, P. (2014). Urban traffic eco-
1325 driving: A macroscopic steady-state analysis. In Control Conference (ECC),
2014 European, pages 2581–2587. IEEE.
41
De Nunzio, G., Canudas-de-Wit, C., Moulin, P., and Di Domenico, D. (2016).
Eco-driving in urban traffic networks using traffic signals information. Inter-
national Journal of Robust and Nonlinear Control, 26(6):1307–1324.
1330 De Nunzio, G., Gomes, G., Canudas-de-Wit, C., Horowitz, R., and Moulin,
P. (2015). Arterial bandwidth maximization via signal offsets and variable
speed limits control. In Decision and Control (CDC), 2015 IEEE 54th Annual
Conference on, pages 5142–5148. IEEE.
De Nunzio, G. and Gutman, P.-O. (2017). An application of shock wave theory
1335 to urban traffic control via dynamic speed advisory. In hEART 2017: 6th
Symposium of the European Association for Research in Transportation.
De Nunzio, G., Thibault, L., and Sciarretta, A. (2017). Bi-objective eco-routing

in large urban road networks. In 2017 IEEE 20th International Conference
on Intelligent Transportation Systems (ITSC), pages 1–7.
1340 Dib, W., Chasse, A., Di Domenico, D., Moulin, P., and Sciarretta, A. (2012).
Evaluation of the energy efficiency of a fleet of electric vehicle for eco-driving
application. Oil & Gas Science and Technology–Revue d’IFP Energies nou-
velles, 67(4):589–599.
Dimitrakopoulos, G. and Demestichas, P. (2010). Intelligent transportation
1345 systems. IEEE Vehicular Technology Magazine, 5(1):77–84.
Dinopoulou, V., Diakaki, C., and Papageorgiou, M. (2006). Applications of the

urban traffic control strategy tuc. European Journal of Operational Research,
175(3):1652 – 1665.
Ericsson, E., Larsson, H., and Brundell-Freij, K. (2006). Optimizing route choice
1350 for lowest fuel consumption – potential effects of a new driver support tool.
Transportation Research Part C: Emerging Technologies, 14(6):369 – 383.
Ferrara, A., Sacone, S., and Siri, S. (2018a). First-order macroscopic traffic
models. In Freeway traffic modelling and control, chapter 3. Springer.
Ferrara, A., Sacone, S., and Siri, S. (2018b). Microscopic and mesoscopic traffic
1355 models. In Freeway traffic modelling and control, chapter 5. Springer.
Ferrara, A., Sacone, S., and Siri, S. (2018c). Second-order macroscopic traffic
Fiori, C., Arcidiacono, V., Fontaras, G., Komnos, D., Makridis, M., Marzano,
V., Mattas, K., Thiel, C., and Ciuffo, B. (2018). Less congestion implies less
1360 energy consumption: Is it really true? Technical report.
Frey, H. C. and Zheng, J. (2002). Probabilistic analysis of driving cycle-
based highway vehicle emission factors. Environmental science & technology,
36(23):5184–5191.
42
Garavello, M., Han, K., and Piccoli, B. (2016). Models for Vehicular Traffic on
1365 Networks.
Gärtner, U., Hohenberg, G., Daudel, H., and Oelschlegel, H. (2004). Devel-
opment and application of a semi-empirical nox model to various hd diesel
engines. In Thermo-and Fluid Dynamic Processes in Diesel Engines 2, pages
285–312. Springer.
1370 Gazis, D. C. and Potts, R. B. (1963). The oversaturated intersection. Technical
report.
Grandinetti, P., Canudas-de Wit, C., and Garin, F. (2018). Distributed optimal
traffic lights design for large-scale urban networks. IEEE Transactions on
Control Systems Technology, (99):1–14.
1375 Greenberg, H. (1959). An analysis of traffic flow. Operations Research, 7(1):79–
85.
Greenshields, B., Channing, W., Miller, H., et al. (1935). A study of traf-
fic capacity. In Highway research board proceedings, volume 1935. National
Research Council (USA), Highway Research Board.
1380 Guanetti, J., Kim, Y., and Borrelli, F. (2018). Control of connected and au-
tomated vehicles: State of the art and future challenges. Annual Reviews in
Control.
Hall, R. (2012). Handbook of transportation science, volume 23. Springer Science

& Business Media.
1385 Han, J., Sciarretta, A., Ojeda, L. L., De Nunzio, G., and Thibault, L. (2018).
Safe-and eco-driving control for connected and automated electric vehicles
using analytical state-constrained optimal solution. IEEE Transactions on
Intelligent Vehicles, 3(2):163–172.
Han, K., Liu, H., Gayah, V. V., Friesz, T. L., and Yao, T. (2016). A robust
1390 optimization approach for dynamic traffic signal control with emission consid-
erations. Transportation Research Part C: Emerging Technologies, 70:3–26.
Hausberger, S. (2009). Emission factors from the model phem for the hbefa
version 3.
Hegyi, A., Hoogendoorn, S. P., Schreuder, M., Stoelhorst, H., and Viti, F.
1395 (2008). Specialist: A dynamic speed limit control algorithm based on shock
wave theory. In 2008 11th International IEEE Conference on Intelligent
Transportation Systems, pages 827–832.
Hegyi, A., Schutter, B. D., and Hellendoorn, H. (2005). Model predictive con-
trol for optimal coordination of ramp metering and variable speed limits.
1400 Transportation Research Part C: Emerging Technologies, 13(3):185 – 209.
43
Helbing, D. (1995). Improved fluid-dynamic model for vehicular traffic. Physical
Review E, 51(4):3164.
Hellström, E., Ivarsson, M., Åslund, J., and Nielsen, L. (2009). Look-ahead
control for heavy trucks to minimize trip time and fuel consumption. Control
1405 Engineering Practice, 17(2):245–254.
Herty, M. and Rascle, M. (2006). Coupling conditions for a class of second-order
models for traffic flow. SIAM Journal on mathematical analysis, 38(2):595–
616.
HomChaudhuri, B., Vahidi, A., and Pisu, P. (2017). Fast model predictive
1410 control-based fuel efficient control strategy for a group of connected vehicles
in urban road conditions. IEEE Transactions on Control Systems Technology,
25(2):760–767.
Hoogendoorn, S. P. and Bovy, P. H. (2001). State-of-the-art of vehicular traffic
flow modelling. Proceedings of the Institution of Mechanical Engineers, Part
1415 I: Journal of Systems and Control Engineering, 215(4):283–303.
Hunt, P., Robertson, D., Bretherton, R., and Winton, R. (1981). Scoot-a traffic
responsive method of coordinating signals. Technical report.
International Energy Agency (2018). World energy outlook 2018.
Jafarmadar, S. (2015). A comparative analysis of two neural network predictions
1420 for performance and emissions in a biodiesel fuelled diesel engine.
Jamshidnejad, A., Papamichail, I., Papageorgiou, M., and De Schutter, B.
(2017). A mesoscopic integrated urban traffic flow-emission model. Trans-
portation Research Part C: Emerging Technologies, 75:45–83.
1425 (2018). Sustainable model-predictive control in urban traffic networks: Effi-
cient solution based on general smoothening methods. IEEE Transactions on
Control Systems Technology, 26(3):813–827.
Jie, L., Zuylen, H. V., Chen, Y., Viti, F., and Wilmink, I. (2013). Calibration
of a microscopic simulation model for emission calculation. Transportation
1430 Research Part C: Emerging Technologies, 31:172 – 184.
Jurik, T., Cela, A., Hamouche, R., Natowicz, R., Reama, A., Niculescu, S.,
and Julien, J. (2014). Energy optimal real-time navigation system. IEEE
Intelligent Transportation Systems Magazine, 6(3):66–79.
Karbowski, D., Kim, N., and Rousseau, A. (2014). Route-based online energy
1435 management of a phev and sensitivity to trip prediction. In Vehicle Power
and Propulsion Conference (VPPC), 2014 IEEE, pages 1–6. IEEE.
Kerner, B. S. and Konhäuser, P. (1993). Cluster effect in initially homogeneous
traffic flow. Physical Review E, 48(4):R2335.
44
Khamis, M. A. and Gomaa, W. (2012). Enhanced multiagent multi-objective
1440 reinforcement learning for urban traffic light control. In 2012 11th Interna-
tional Conference on Machine Learning and Applications, volume 1, pages
586–591. IEEE.
Khondaker, B. and Kattan, L. (2015a). Variable speed limit: A microscopic
analysis in a connected vehicle environment. Transportation Research Part
1445 C: Emerging Technologies, 58:146 – 159.
Khondaker, B. and Kattan, L. (2015b). Variable speed limit: An overview.
Transportation Letters, 7.
Kluge, S., Santa, C., Dangl, S., Wild, S., Brokate, M., Reif, K., and Busch,
F. (2013). On the computation of the energy-optimal route dependent on
1450 the traffic load in ingolstadt. Transportation Research Part C: Emerging
Kubička, M., Klusáček, J., Sciarretta, A., Cela, A., Mounier, H., Thibault, L.,
and Niculescu, S.-I. (2016). Performance of current eco-routing methods. In
Intelligent Vehicles Symposium (IV), 2016 IEEE, pages 472–477. IEEE.
1455 Lejri, D., Can, A., Schiper, N., and Leclercq, L. (2018). Accounting for traffic
speed dynamics when calculating copert and phem pollutant emissions at the
urban scale. Transportation Research Part D: Transport and Environment,
63:588 – 603.
Lelouvier, A., Guanetti, J., and Borrelli, F. (2017). Eco-platooning of au-

1460 tonomous electrical vehicles using distributed model predictive control. pa-
rameters, 2:4.
Lighthill, M. J. and Whitham, G. B. (1955). On kinematic waves ii. a theory of
traffic flow on long crowded roads. Proc. R. Soc. Lond. A, 229(1178):317–345.
Lin, S., De Schutter, B., Xi, Y., and Hellendoorn, H. (2012). Efficient network-
1465 wide model-based predictive control for urban traffic networks. Transportation
Research Part C: Emerging Technologies, 24:122–140.
Lin, S., De Schutter, B., Xi, Y., and Hellendoorn, H. (2013). Integrated ur-
ban traffic control for the reduction of travel delays and emissions. IEEE
Transactions on Intelligent Transportation Systems, 14(4):1609–1619.
1470 Lin, S. and Xi, Y. (2008). An efficient model for urban traffic network control.
IFAC Proceedings Volumes, 41(2):14066–14071.
Liu, F., Zhao, F., Liu, Z., and Hao, H. (2019). Can autonomous vehicle reduce
greenhouse gas emissions? a country-level evaluation. Energy Policy, 132:462
– 473.
1475 Liu, R. and Tate, J. (2004). Network effects of intelligent speed adaptation
systems. Transportation, 31(3):297–325.
45
Liu, S., Hellendoorn, H., and De Schutter, B. (2017). Model predictive control
for freeway networks based on multi-class traffic flow and emission models.
IEEE Transactions on Intelligent Transportation Systems, 18(2):306–320.
1480 Lowrie, P. et al. (1982). The sydney co-ordinated adaptive traffic system (scats)
– principles, methodology, algorithms. In Proceedings of the International
Conference on Road Traffic Signaling, London, UK.
Luo, L., Ge, Y.-E., Zhang, F., and Ban, X. J. (2016). Real-time route diver-
sion control in a model predictive control framework with multiple objectives:
1485 Traffic efficiency, emission reduction and fuel economy. Transportation Re-
search Part D: Transport and Environment, 48:332–356.
Mahler, G. and Vahidi, A. (2012). Reducing idling at red lights based on prob-
abilistic prediction of traffic signal timings. In American Control Conference
(ACC), 2012, pages 6557–6562. IEEE.
1490 Mandl, N. and Pinterits, M. (2018). Annual european union greenhouse gas
inventory 1990–2016 and inventory report 2018.
Mehr, N., Sanselme, M., Orr, N., Horowitz, R., and Gomes, G. (2018). Offset
selection for bandwidth maximization on multiple routes. In 2018 Annual
American Control Conference (ACC), pages 6366–6371. IEEE.
1495 Messmer, A. and Papageorgiou, M. (1990). Metanet: a macroscopic simulation
program for motorway networks. Traffic Engineering & Control, 31:466–470.
Mirchandani, P. and Head, L. (2001). A real-time traffic signal control sys-
tem: architecture, algorithms, and analysis. Transportation Research Part C:
Emerging Technologies, 9(6):415 – 432.
1500 Miyatake, M., Kuriyama, M., and Takeda, Y. (2011). Theoretical study on eco-
driving technique for an electric vehicle considering traffic signals. In Power
Electronics and Drive Systems (PEDS), 2011 IEEE Ninth International Con-
ference on, pages 733–738. IEEE.
Möhner, M. (2018). Driving ban for diesel-powered vehicles in major cities: an
1505 appropriate penalty for exceeding the limit value for nitrogen dioxide?
Nannicini, G., Delling, D., Schultes, D., and Liberti, L. (2012). Bidirectional a*
search on time-dependent road networks. Networks, 59(2):240–251.
Negrenti, E. (1999). The ‘corrected average speed’ approach in enea’s tee model:
an innovative solution for the evaluation of the energetic and environmental
1510 impacts of urban transport policies. Science of the total environment, 235(1-
3):411–413.
Newell, G. F. (1993). A simplified theory of kinematic waves in highway traf-
fic, part i: General theory. Transportation Research Part B: Methodological,
27(4):281–287.
46
1515 Ntziachristos, L., Gkatzoflias, D., Kouridis, C., and Samaras, Z. (2009). Copert:
A european road transport emission inventory model. pages 491–504.
Obodeh, O. and Ajuwa, C. (2009). Evaluation of artificial neural network per-
formance in predicting diesel engine nox emissions. European Journal of Sci-
entific Research, 33(4):642–653.
1520 Osorio, C. and Nanduri, K. (2015). Energy-efficient urban traffic management:

a microscopic simulation-based approach. Transportation Science, 49(3):637–
651.
Ozatay, E., Onori, S., Wollaeger, J., Özgüner, Ü., Rizzoni, G., Filev, D., Miche-
lini, J., and Cairano, S. D. (2014a). Cloud-based velocity profile optimization
1525 for everyday driving: A dynamic-programming-based solution. IEEE Trans-
actions on Intelligent Transportation Systems, 15:2491–2505.
Ozatay, E., Ozguner, U., Michelini, J., and Filev, D. (2014b). Analytical solu-
tion to the minimum energy consumption based velocity profile optimization
problem with variable road grade. IFAC Proceedings Volumes, 47(3):7541–
1530 7546.
Panis, L. I., Broekx, S., and Liu, R. (2006). Modelling instantaneous traffic
emission and the influence of traffic speed limits. Science of the total envi-
ronment, 371(1-3):270–285.
Papageorgiou, M., Hadj-Salem, H., Blosseville, J.-M., et al. (1991). Alinea: A

1535 local feedback control law for on-ramp metering. Transportation Research
Record, 1320(1):58–67.
Pasquale, C., Papamichail, I., Roncoli, C., Sacone, S., Siri, S., and Papageor-
giou, M. (2015). Two-class freeway traffic regulation to reduce congestion
and emissions via nonlinear optimal control. Transportation Research Part
1540 C: Emerging Technologies, 55:85–99.
Payne, H. (1971). Models of freeway traffic and control. In Mathematical Models
of Public Systems. Simul. Counc. Proc.
Pelkmans, L. and Debal, P. (2006). Comparison of on-road emissions with emis-
sions measured on chassis dynamometer test cycles. Transportation Research
1545 Part D: Transport and Environment, 11(4):233 – 241.
Piccoli, B. and Tosin, A. (2009). A review of continuum mathematical models
of vehicular traffic. Encyclopedia of Complexity and Systems Science, pages
9727–9749.
Pisarski, D. and Canudas-de-Wit, C. (2016). Nash game-based distributed con-

1550 trol design for balancing traffic density over freeway networks. IEEE Trans.
Control of Network Systems, 3(2):149–161.
47
Post, K., Kent, J., Tomlin, J., and Carruthers, N. (1984). Fuel consumption and
emission modelling by power demand and a comparison with other models.
Transportation Research Part A: General, 18(3):191–213.
1555 Qi, Y. G., Teng, H. H., and Yu, L. (2004). Microscale emission models incorpo-
rating acceleration and deceleration. Journal of Transportation Engineering,
130(3):348–359.
Qu, L., Li, L., Zhang, Y., and Hu, J. (2009). Ppca-based missing data imputa-
tion for traffic flow volume: A systematical approach. IEEE Transactions on
1560 intelligent transportation systems, 10(3):512–522.
Rai, N. K., Vyas, A., and Singh, S. K. (2017). Street level modeling of pollu-
tants for residential areas. International Journal of Engineering Research and
Application, 7(10):52–57.
Rakha, H., Ahn, K., and Trani, A. (2003). Comparison of mobile5a, mobile6,
1565 vt-micro, and cmem models for estimating hot-stabilized light-duty gasoline
vehicle emissions. Canadian Journal of Civil Engineering - CAN J CIVIL
ENG, 30:1010–1021.
Richards, P. I. (1956). Shock waves on the highway. Operations research,
4(1):42–51.
1570 Rios-Torres, J. and Malikopoulos, A. A. (2018). Impact of partial penetrations
of connected and automated vehicles on fuel consumption and traffic flow.
IEEE Transactions on Intelligent Vehicles, 3(4):453–462.
Schiper, N. (2017). Traffic data sampling for air pollution estimation at different
urban scales. PhD thesis, Université de Lyon.
1575 Sciarretta, A., Nunzio, G. D., and Ojeda, L. L. (2015). Optimal ecodriving con-
trol: Energy-efficient driving of road vehicles as an optimal control problem.
IEEE Control Systems Magazine, 35(5):71–90.
Scora, G. and Barth, M. (2006). Comprehensive modal emissions model (cmem),
version 3.01. User guide. Centre for Environmental Research and Technology.
1580 University of California, Riverside, 1070.
Seredynski, M., Mazurczyk, W., and Khadraoui, D. (2013). Multi-segment green

light optimal speed advisory. In Parallel and Distributed Processing Sympo-
sium Workshops & PhD Forum (IPDPSW), 2013 IEEE 27th International,
pages 459–465. IEEE.
1585 Sobrino, N., Monzón, A., and Hernandez, S. (2014). Reduced carbon and en-
ergy footprint in highway operations: The highway energy assessment (hera)
methodology. Netw Spat Econ, 16:1–20.
48
Stern, R. E., Chen, Y., Churchill, M., Wu, F., Monache, M. L. D., Piccoli, B.,
Seibold, B., Sprinkle, J., and Work, D. B. (2019). Quantifying air quality ben-
1590 efits resulting from few autonomous vehicles stabilizing traffic. Transportation
Research Part D: Transport and Environment, 67:351 – 365.
Stevanovic, A., Stevanovic, J., Zhang, K., and Batterman, S. (2009). Optimizing
traffic control to reduce fuel consumption and vehicular emissions: Integrated
approach with vissim, cmem, and visgaost. Transportation Research Record,
1595 2128(1):105–113.
Sturm, P., Kirchweger, G., Hausberger, S., and Almbauer, R. (1998). Instanta-
neous emission data and their use in estimating road traffic emissions. Inter-
national Journal of Vehicle Design, 20.
Sun, C., Guanetti, J., Borrelli, F., and Moura, S. (2018). Robust eco-driving
1600 control of autonomous vehicles connected to traffic lights. arXiv preprint
arXiv:1802.05815.
Sun, J. and Liu, H. X. (2015). Stochastic eco-routing in a signalized traffic

network. Transportation Research Part C: Emerging Technologies, 59:32 –
47. Special Issue on International Symposium on Transportation and Traffic
1605 Theory.
Tajali, M. and Hajbabaie, A. (2018). Dynamic speed harmonization in urban
street networks. Computer-Aided Civil and Infrastructure Engineering.
Tampère, C. M., Corthout, R., Cattrysse, D., and Immers, L. H. (2011). A

generic class of first order node models for dynamic macroscopic simulation
1610 of traffic flows. Transportation Research Part B: Methodological, 45(1):289 –
309.
Taylor, M. (2000). Network modeling of the traffic, environmental and energy

effects of lower urban speed limits. Road and Transport Research, 9:48–57.
Thibault, L., Degeilh, P., Lepreux, O., Voise, L., Alix, G., and Corde, G. (2016).
1615 A new gps-based method to estimate real driving emissions. In 2016 IEEE
19th International Conference on Intelligent Transportation Systems (ITSC),
pages 1628–1633.
Transport & Environment (2018). Co2 emissions from cars: the facts.
Treiber, M. and Kesting, A. (2013a). Model-based traffic flow optimization. In
1620 Traffic Flow Dynamics, chapter 21. Springer.
Treiber, M. and Kesting, A. (2013b). Trajectory and floating-car data. In Traffic

Flow Dynamics, chapter 2. Springer.
U.S. Department of Energy (2018). Autonomous vehicles: Uncertainties and
energy implications.
49
1625 U.S. Energy Information Administration (2017a). International energy outlook.
U.S. Energy Information Administration (2017b). Study of the potential energy
consumption impacts of connected and automated vehicles.
Van den Berg, M., Hegyi, A., De Schutter, B., and Hellendoorn, H. (2007).
Integrated traffic control for mixed urban and freeway networks: A model
1630 predictive control approach. European journal of transport and infrastructure
research EJTIR, 7 (3).
van Wageningen-Kessels, F., van Lint, H., Vuik, K., and Hoogendoorn, S.
(2015). Genealogy of traffic flow models. EURO Journal on Transportation
and Logistics, 4(4):445–473.
1635 Varaiya, P. (2013). The max-pressure controller for arbitrary networks of sig-
nalized intersections. In Advances in Dynamic Network Modeling in Complex
Transportation Systems, pages 27–66. Springer.
Vreeswijk, J. D., Mahmod, M., and van Arem, B. (2013). Energy efficient traffic
management and control - the ecomove approach and expected benefits. In
1640 13th International IEEE Conference on Intelligent Transportation Systems,
pages 955–961.
Walraven, E., Spaan, M. T., and Bakker, B. (2016). Traffic flow optimization:
A reinforcement learning approach. Engineering Applications of Artificial
Intelligence, 52:203 – 212.
1645 Wang, Y., Szeto, W. Y., Han, K., and Friesz, T. L. (2018). Dynamic traffic
assignment: A review of the methodological advances for environmentally
sustainable road transportation applications. Transportation Research Part
B: Methodological, 111:370–394.
Whitham, G. B. (1974). Linear and nonlinear waves. John Wiley & Sons.
1650 World Health Organization (2016). Ambient air pollution: a global assessment
of exposure and burden of disease.
Wu, C., Bayen, A. M., and Mehta, A. (2018). Stabilizing traffic with autonomous
vehicles. In 2018 IEEE International Conference on Robotics and Automation
(ICRA), pages 1–7. IEEE.
1655 Wu, G., Boriboonsomsin, K., and Barth, M. J. (2014). Development and evalu-
ation of an intelligent energy-management strategy for plug-in hybrid electric
vehicles. IEEE Trans. Intelligent Transportation Systems, 15(3):1091–1100.
Xie, B., Xu, M., Härri, J., and Chen, Y. (2013). A traffic light extension to
cell transmission model for estimating urban traffic jam. In PIMRC, pages
1660 2566–2570.
50
Xu, T. D., Sun, L. J., Peng, Z. R., and Hao, Y. (2011). Integrated route guidance
and ramp metering consistent with drivers’ en-route diversion behaviour. IET
Intelligent Transport Systems, 5(4):267–276.
Xu, Z., Kang, Y., and Lv, W. (2017). Analysis and prediction of vehicle exhaust
1665 emission using ann. In 2017 36th Chinese Control Conference (CCC), pages
4029–4033. IEEE.
Yang, H. and Jin, W.-L. (2014). A control theoretic formulation of green driving
strategies based on inter-vehicle communications. Transportation Research
Part C: Emerging Technologies, 41:48–60.
1670 Yperman, I. (2007). The link transmission model for dynamic network loading.
Zegeye, S., De Schutter, B., Hellendoorn, J., Breunesse, E., and Hegyi, A.
(2013). Integrated macroscopic traffic flow, emission, and fuel consumption
model for control purposes. Transportation Research Part C: Emerging Tech-
nologies, 31:158–171.
1675 Zegeye, S. K. (2011). Model-based traffic control for sustainable mobility.

Zegeye, S. K., De Schutter, B., Hellendoorn, J., Breunesse, E. A., and Hegyi,
A. (2012). A predictive traffic controller for sustainable mobility using pa-
rameterized control policies. IEEE Transactions on Intelligent Transportation
Systems, 13(3):1420–1429.
1680 Zhang, H. M. (2002). A non-equilibrium traffic model devoid of gas-like behav-

ior. Transportation Research Part B: Methodological, 36(3):275–290.
Zhang, Y. and Cassandras, C. G. (2018). The penetration effect of connected
automated vehicles in urban traffic: An energy impact study. arXiv preprint
arXiv:1803.05577.
1685 Zhu, F. and Ukkusuri, S. V. (2014). Accounting for dynamic speed limit con-
trol in a stochastic traffic environment: A reinforcement learning approach.
Transportation Research Part C: Emerging Technologies, 41:30 – 47.
Zu, Y., Liu, C., Dai, R., Sharma, A., and Dong, J. (2018). Real-time energy-
efficient traffic control via convex optimization. Transportation Research Part
51
Version of Record: https://www.sciencedirect.com/science/article/pii/S1367578819300446
Manuscript_e3963856058ab91a75c58b3dfe2045fe
Ecological traffic management: a review of the

modeling and control strategies for improving
environmental sustainability of road transportation
Bassel Othmana,b,∗, Giovanni De Nunzioa , Domenico Di Domenicoa , Carlos

Canudas-de-Witb
a IFP Energies nouvelles, Rond-point de l’échangeur de Solaize, BP 3, 69360 Solaize, France
b Univ. Grenoble Alpes, CNRS, Inria, Grenoble INP, GIPSA-lab, 38000 Grenoble, France
Abstract
As road transportation energy use and environmental impact are globally rising
at an alarming pace, authorities seek in research and technological advancement
innovative solutions to increase road traffic sustainability. The unclear and par-
tial correlation between road congestion and environmental impact is promoting
new research directions in traffic management. This paper aims to review the
existing modeling approaches to accurately represent traffic behavior and the
associated energy consumption and pollutant emissions. The review then covers
the transportation problems and control strategies that address directly environ-
mental performance criteria, especially in urban networks. A discussion on the
advantages of the different methods and on the future outlook for the eco-traffic
management completes the proposed survey.
Keywords: traffic modeling, traffic management, energy efficiency, pollutant
emissions, optimization
1. Introduction
While energy-related air pollution is considered today one of the primary

premature death causes (World Health Organization, 2016), the global carbon
dioxide (CO2 ) emissions are on a rising trend destined to grow well above the
5 levels imposed by the international climate goals (International Energy Agency,
2018). Population surge and economic growth of the developing countries have
been identified as the main causes of the drastic increase of energy demand and
pollutant emissions in all sectors (International Energy Agency, 2018).
E-mail addresses: bassel.othman@ifpen.fr (B. Othman), giovanni.de-nunzio@ifpen.fr

(G. De Nunzio), carlos.canudas-de-wit@gipsa-lab.fr (C. Canudas de Wit),
domenico.didomenico@ifpen.fr (D. Di Domenico).
Preprint submitted to Annual Reviews in Control June 28, 2019
© 2019 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
The worldwide transportation sector alone accounts for 55% of the total liq-
10 uid fuels consumption and, with the increasing travel demand, this share is not
expected to decrease for the next two decades (U.S. Energy Information Admin-
istration, 2017a). In the member countries of the Organisation for Economic
Co-operation and Development (OECD), projections show that the improved
energy efficiency in transportation may lead to a net decline of about 2% in
15 energy use until 2040, thus outpacing the predicted increase of vehicle-miles
traveled (VMT). However, in OECD-Europe, transportation still represents the
biggest source of carbon emissions (Transport & Environment, 2018), contribut-
ing by about 25% to the total CO2 emissions, with cars and vans representing
more than two thirds of this share (Mandl and Pinterits, 2018). The situation
20 is even more alarming in non-OECD countries, where the transportation energy
demand is expected to rise by 64% until 2040, implying an increase of about
15% of energy-related CO2 emissions (U.S. Energy Information Administration,
2017a).
Therefore, a lot of attention has been drawn worldwide to finding the most ef-
25 fective measures to help reduce the current contribution to greenhouse gas emis-
sions from transportation. Governments, practitioners and researchers seem to
agree on the fact that a combination of short-term and long-term strategies must
be adopted. In the short-term, policies and regulations encouraging changes in
behavior and travel habits represent a key lever. Attractiveness of alternative
30 means of transportation should be enhanced, a shift to less polluting transport
modes should be promoted, and a change in purchasing habits favoring smaller
and more energy-efficient cars should be encouraged (Chapman, 2007). In the
long-term, the widespread adoption of innovative technological solutions such as
electrification, connectivity and automation are expected to enable a significant
35 shift in the future of personal transportation and mobility. The way for such
a technological transformation of mobility is already being paved thanks to the
diffusion of connected and automated vehicles (CAVs), multi-vehicle (V2V) and
vehicle-infrastructure (V2I) cooperation and communication networks, in- and
over-roadway sensors, cloud-computing capabilities, etc. (Guanetti et al., 2018).
40 However, the potential energy benefits of these technologies remain uncer-
tain, mostly because of the high level of non-linear dependence between different
aspects of an automated transportation system operating with conventional ve-
hicles, as well as possible side-effects of automation (U.S. Energy Information
Administration, 2017b). Among the features enabled by the aforementioned
45 technologies that promise to increase energy efficiency and reduce pollutant
emissions of transportation, it is worth mentioning eco-driving, eco-routing,
platooning, roadway throughput optimization, powertrain electrification, vehi-
cle down-sizing, parking search time reduction, ride-sharing. On the other hand,
as for the side-effects that may endanger energy efficiency and emission reduc-
50 tion, it is likely that technology may increase traffic congestion as a consequence
of an increased access to mobility, increase travel speeds as a consequence of en-
hanced safety, increase commute distances as an effect of increased comfort and
reduced travel costs, etc. (U.S. Energy Information Administration, 2017b).
From a single-vehicle efficiency perspective, research suggests that lightweight,
2
55 low-speed, autonomous vehicles have the potential to achieve fuel economies an
order of magnitude higher than current cars (U.S. Energy Information Admin-
istration, 2017b). However, at system-wide level, current estimates suggest that
the total energy consumption impacts can range from a 90% decrease to a 200%
increase in fuel consumption as compared to a projected 2050 baseline energy
60 (Brown et al., 2014).
Such a large variability in the possible outcome of the adoption of the new
vehicular and traffic technologies makes it somewhat difficult to focus and pri-
oritize the research efforts to increase energy efficiency of mobility. Nowadays,
the general trend in research and policy seems to aim to reduce CO2 emissions
65 by pushing for more efficient vehicles and reducing VMT. This is based on a
generally accepted paradigm that congestion mitigation programs should reduce
CO2 emissions. However, it is difficult to prove a clear direct proportionality
between congestion and CO2 emissions (Fiori et al., 2018). The most reliable
approach to improve energy efficiency and reduce pollutant emissions in the de-
70 sign of a traffic regulation measure consists in directly considering these aspects
as decision and optimization criteria. Therefore, interest in transportation reg-
ulation problems with explicit environmental considerations is growing (Wang
et al., 2018; Vreeswijk et al., 2013).
This paper surveys the existing scientific literature on energy consumption
75 and emission models, as well as road transportation problems directly address-
ing the issue of energy consumption and pollutant emissions reduction. Such
problems can be tackled at different levels depending on the granularity and
the object of the control action. At vehicle level, the energy-efficient control
strategies typically act on single vehicles or groups of cooperating vehicles by
80 modifying their individual speed profiles or route choices. At traffic level, the
control strategies aim to influence the vehicular flow as a whole by acting on
the typical flow regulation actuators, such as traffic lights, speed limits, etc.
The adopted categorization in terms of modeling and control approaches both
at vehicle and traffic level for the general problem of reducing environmental
85 impact of road transportation is illustrated in Fig. 1.
The contributions of this paper are summarized as follows:
• A comprehensive literature review of the existing energy consumption and
pollutant emissions models is provided. The review distinguishes between
data and physics-based models and discusses their adaptation for usage
90 with both single-vehicles and traffic flow.
• An overview of the existing vehicle and traffic control strategies to im-
prove energy and environmental efficiency of transportation is given. The
review focuses on the control techniques that explicitly address energy con-
sumption and emissions. The connection and interaction between traffic
95 congestion and energy efficiency is also discussed.
• As an outcome of this review, research gaps in the current state of the art
have been identified and discussed in order to inspire future works in this
field.
3
Single vehicle Emission/energy
Kinematics Control
models
Speed Data-driven Eco-driving

measurements
Microscopic Physical Eco-routing

traffic models
Emission/energy
Traffic Kinematics Control
models
Speed Traffic
Data-driven lights
measurements
Speed
Fluid-based Physical limits
models
Dynamic
routing
Automated
vehicles
Figure 1: Diagram of the global approach for energy consumption and emissions modeling
and control for single vehicles and traffic flow.
The body of the paper is organized as follows. Section II presents the energy
100 consumption and emission models for the single vehicle with a brief discussion of
how the vehicle kinematics can be obtained. Analogously, Section III introduces
the modeling approaches to describe traffic kinematics, with a particular focus
on the most popular fluid-dynamics traffic models, as well as the energy con-
sumption and emission models for vehicular flow. The energy-optimal control
105 strategies for single vehicles are presented in Section IV, while the transporta-
tion problems dealing with traffic energy efficiency are reviewed in Section V.
Finally, Section VI contains concluding remarks and discussion on the current
research gaps and future outlooks.
2. Emission and energy consumption models for single vehicles
110 Different models estimating emissions and energy consumption rate (Jy ) of
a vehicle as a function of its parameters and operation variables (u) have been
investigated in the past. This section presents the data-driven and the physical
modeling approaches employed to estimate Jy .
In the proposed formalization, Jy refers to the prediction of the rate of y,
which can be calculated per distance traveled by the vehicle (Jyspat ) or per time
unit (Jytemp ), depending on the modeling method. y corresponds either to the
emission of a pollutant (CO, N Ox , HC, ...) or the energy consumption (fuel or
4
Microscopic
emission/energy models
Data-driven Physical modeling

approach approach
Look-up table Regression Deterministic Probabilistic

models models methods methods
Figure 2: Classification of emission and energy consumption models for single vehicles.
electricity consumption, depending on the vehicle powertrain considered):

y ∈ {fuel or electricity consumption, emission of CO, N Ox , HC, ...}
Such emission and energy consumption models are said microscopic because
115 they consider each vehicle individually. They can be described as
Jy = f (u) (1)
where f is a function that relates the model inputs to the output.
The function f can be constructed in different ways. The different ap-
proaches detailed in this section to estimate emissions and energy consumption
are classified as illustrated in Fig. 2.
120 The first step to determine the emission and energy consumption rates of
a vehicle is to determine its operation variables (e.g. speed, acceleration). A
solution is to obtain these data by sensors. For example, Thibault et al. (2016)
propose to use smartphone devices and their embedded sensors to get the posi-
tion and speed of vehicles. Treiber and Kesting (2013b) present a methodology
125 to express the operation variables of vehicles from trajectory and floating-car
data.
They can also be determined through simulation using a microscopic traffic
model, which reproduces the movement of each vehicle individually. Some com-
plete overviews of microscopic traffic models can be found in van Wageningen-
130 Kessels et al. (2015); Ferrara et al. (2018b); Hoogendoorn and Bovy (2001).
These approaches are mainly based on the car-following principle (e.g. safe-
distance models, stimulus-response models, action point models). For example,
the optimal velocity car-following model expresses the acceleration of each ve-
hicle as
ve − v
a= (2)
τ
135 where the optimal speed, ve , depends on the distance with the vehicle upstream,
and τ is the driver reaction time.
The use of a microscopic traffic model, especially in order to estimate emis-
sions and energy consumption, requires a precise calibration of model param-
eters. Jie et al. (2013) present a methodology to perform such a parameters
5
a Vehicle acceleration [m.s−2 ]
A Cross-sectional area [m2 ]
by Stoichiometric CPF (Catalyst Pass Fraction)
coefficient [s.kg−1 ]
cy Enrichment CPF coefficient [s.kg−1 ]
Ca Aerodynamic drag coefficient [-]
Cd Reynolds coefficient [-]
Ce Engine friction factor [J.rev−1 .m−3 ]
Cr Rolling resistance coefficient [-]
COC Center of combustion (50% energy conversion, from
Top Dead Center) [crank angle degree]
CP Fy Catalyst pass fraction of y [-]
d Mass density of air [kg.m−3 ]
D Engine displacement [m3 ]
g Gravitational constant [m.s−2 ]
LHVfuel Fuel lower heating value [J.kg−1 ]
mcyl In-cylinder air mass per stroke and displaced
mO2 In-cylinder oxygen mass per stroke and displaced
M Vehicle mass [kg]
n Engine speed [revolutions per minute]
Pacc Engine power demand associated with accessories [W]
RBGR In-cylinder burnt gas ratio [-]
v Vehicle speed [m.s−1 ]
y Maximum catalyst efficiency of y [-]
α Road grade angle [rad]
λ Ratio between the air/fuel ratio at stochiometry and
the commanded air/fuel ratio [-]
ηbatt Battery efficiency [-]
ηeng Engine efficiency [-]
ηtf Efficiency of the transmission and final drive [-]
Table 1: Parameters and operation variables of vehicles used in the emission and energy
consumption models.
140 calibration and emphasize on its benefits in terms of speed and acceleration
estimation.
The second step to determine the emission and energy consumption rates
of a vehicle is to use a microscopic emission and energy consumption model
whose inputs are the vehicle operating variables and parameters, summarized
145 in Table 1. This step is presented in detail in the following sections.
6
2.1. Data-driven methods
Emission and energy consumption rates can be calculated using data-driven
approaches. These can be either based on look-up table models or regression
models.
150 2.1.1. Look-up table models

One old common approach to estimate emission and fuel consumption rates
consists in performing chassis dynamometer tests and recording the emissions
and fuel consumption in a look-up table, also called emission matrix. Usually,
such look-up tables provide Jy from speed and acceleration (Post et al., 1984;
155 Sturm et al., 1998) for a given set of vehicle parameters. These reference emis-
sion look-up tables can be used later to instantly estimate emissions and fuel
consumption.
Although this method is easy to use, usually the available matrices are
sparse, due to measurement difficulties. Moreover, empirical matrix-based pre-
160 diction concerns only steady-state emissions, and not transient operations (Scora
and Barth, 2006). Finally, this method is sensitive to the driving cycle and the
quality of on-line measurements. This may lead to large errors.
Another possibility is to determine emission and fuel consumption rates as a
function of the vehicle position. Andersen et al. (2013) propose to associate to
165 each road a corresponding fuel consumption, based on average measurements.
The amount of fuel consumed by a vehicle during a trip is therefore simply
approximated by the sum of the average fuel consumption associated with the
corresponding roads. This approach is very simple but it cannot distinguish
between different types of vehicle, as they are all mixed in the same computed
170 average value. Also, it cannot reflect the evolution of emissions and fuel con-
sumption in case of traffic congestion.
2.1.2. Regression models

Emissions and energy consumption of a single vehicle can also be predicted
on a second-by-second basis by using data-based models, such as regression
175 techniques or neural networks. The inputs of these models can typically be
the speed, acceleration or power demand, and the outputs are the emission or
energy consumption rates prediction.
Regression techniques and neural networks for emission and energy consump-
tion modeling both use the collected data in order to train a model that mimics
180 these data. In regression techniques, it is necessary to identify the model pa-
rameters by curve fitting, while in neural networks the weight of the connections
between neurons is to be identified.
The use of neural networks to estimate emissions and energy consumption
is motivated by the heavy nonlinearity of emissions. There is also a need of
185 high computational efficiency in order to be compatible with second-by-second
microscopic traffic models. Such neural network frameworks can be found in
Ahn (1998); Obodeh and Ajuwa (2009); Jafarmadar (2015); Xu et al. (2017).
Ahn (1998) presents non-linear multiple regression models constructed with
quadratic and cubic speed-acceleration terms. The data used to determine the
7
190 coefficients of these models for a given type of vehicle is obtained from dy-
namometer emission tests, based on the New European Drive Cycle (NEDC)
(Ahn et al., 2004). It is also desirable to use data from vehicles in real urban
traffic situations, when available (Panis et al., 2006). In fact, it is important to
note that emission levels obtained from dynamometer tests can be much lower
195 than those produced in real traffic (Pelkmans and Debal, 2006). For example,
a criticism against the NEDC is that its acceleration profile is very smooth and
not sufficiently realistic (Andre and Pronello, 1997).
Based on this technique, the VT-micro model can be formulated in matrix
form (Zegeye et al., 2013) as
ln(Jytemp ) = ~v My~a (3)
200 where My denotes the regression coefficients matrix of y for the type of vehicle
under consideration, ~v and ~a are respectively the speed and acceleration vectors
defined as
~v = 1, v, v 2 , v 3

T (4)
~a = 1, a, a2 , a3

Note that the VT-micro model can also be expressed with a regression co-
efficients matrix for positive accelerations, and another matrix for negative ac-
205 celerations, depending on the data used to calibrate the model (Alsabaan et al.,
2012). VT-micro estimates emissions and energy consumption from instanta-
neous speed and acceleration, i.e. measured at the present time. Qi et al. (2004)
formulate a regression model, named POLY, which also takes into account the
past accelerations and the road grade angle. The model reads
Jytemp = β0 + β1 v(k) + β2 v 2 (k) + β3 v 3 (k)
+β4 T acc (k) + β5 T dec (k) (5)
+β6 ga (k) + ... + β15 ga (k − 9) + β16 v(k)ga (k)

210 where β0 to β16 are the parameters determined by least-square method for
one type of vehicle, T acc (k) and T dec (k) are respectively the acceleration and
deceleration duration since their inception up to the current time step k. At
each time step, at least one of them is zero. To consider the grade angle α, the
function ga is defined as follows
" #
α(k)
ga (k) = a(k) + g p (6)
1 + α2 (k)
215 POLY is an accurate emission model. However, it may underestimate emis-
sions of higher emitting vehicles as it is built from average measured data (Qi
et al., 2004).
While data-driven models can be developed quickly without prior knowledge
on the vehicle or roads, they usually lack a clear physical interpretation and
220 might be too coarse. They may also over-fit the calibration data if the number
of variables considered is too large.
8
2.2. Physical modeling approach
An alternative method for estimating emissions and energy consumption is
to employ a physical approach that leads to model parameters with physical
225 meaning. Two types of models can be distinguished, the deterministic and
probabilistic models, that are both described in the following sections.
2.2.1. Deterministic methods

The emission and energy consumption rates can be determined from the
power engine demand P , which can be calculated using the following vehicle
230 longitudinal dynamics, as in Sciarretta et al. (2015)

M a = Ftrac − Fb − Fres
(7)
a = dv
dt
where Ftrac is the traction force transmitted by the powertrain to the wheels,
Fb is the mechanical brake force and Fres is the resistance force that can be
calculated as follows
1
Fres = M g sin α + M gCr + dv 2 ACa (8)
2
The total tractive power of the vehicle is denoted Ptrac and is given by
Ptrac = Ftrac v (9)
235 (7) to (9) lead to

d
Ptrac = M v (a + g sin α) + v M gCr + v 2 ACa + Fb v (10)
2
Finally, the power engine demand P can be calculated as follows
Ptrac
P = + Pacc (11)
ηtf
Once the power demand is known, Post et al. (1984) propose to estimate the
emission and energy consumption rates as follows

ay + by P , if P ≥ 0
Jytemp = (12)
ay , if P < 0
ay and by are the regression coefficients determined for a given y and vehicle
240 type. When y represents the fuel consumption, ay can be approximated by the
following linear function
ay = γ × D (13)
where γ is a constant.
Barth et al. (1996) propose to replace the regression coefficients by physical
parameters and operation variables to approximate the fuel use rate. The model
245 is defined as
temp P
Jf uel LHVf uel ≈ λ Ce nD + (14)
ηeng
9
where the engine friction factor Ce is the energy used at zero-power output to
overcome engine friction.
An et al. (1997) then propose to calculate the pollutant emission rates as
follows
dy
Jytemp = Jftemp
uel CP Fy (15)
d(f uel)
250 where y refers here only to emissions, d(fdy
uel) corresponds to the grams of engine-
out emissions per gram of fuel consumed for pollutant y, and the catalyst pass
fraction CP Fy can be modeled as

1
CP Fy = 1 − y exp −by − cy 1 − Jftemp
uel (16)
λ
The Comprehensive Modal Emissions Model (CMEM) is based on (15)
(Scora and Barth, 2006). It considers different categories of vehicle and different
255 modes of operation (idling, cruising, acceleration, and deceleration). Emission
and fuel consumption rates are calculated as a function of the vehicle fleet com-
position (vehicle categorization based on model year, weight, etc.), operation
variables and model-calibrated parameters. The structure of the model is shown
in Fig. 3. The CMEM predicts emissions well, but may underestimate them
260 for high-emitting vehicles because the model is based on the average data of
300 vehicles (including about 30 high emitters) measured during dynamometer
tests, along different driving cycles (Rakha et al., 2003).
Figure 3: Structure of the CMEM (Scora and Barth, 2006) [Published with permission of the
Center for Environmental Research and Technology].
Another model proposed by Gärtner et al. (2004) estimates emissions from

fuel consumption at engine mechanics level. The model relies on the first Law of
10
265 Thermodynamics and chemical kinetic reaction rate considerations. For N Ox
emissions, the model reads

d(N Ox )
log = a0 + a1 COC + a2 mcyl + a3 mO2 (17)
d(f uel)
where d(N Ox )
d(f uel) is the mass of nitrogen oxides emitted per mass of fuel consumed,
and a0 , a1 , a2 , a3 are model coefficients.
A simplified version of (17) is proposed by Thibault et al. (2016). The
270 authors propose to express log(JN Ox ) as a linear function of the in-cylinder
burnt gas ratio RBGR , as follows

d(N Ox )
log = a4 + a5 RBGR (18)
d(f uel)
where a4 , a5 are model coefficients.

RBGR is expressed as a function of the engine speed and the engine torque
for a given type of vehicle, based on the data from the NEDC. The engine con-
275 ditions are physically determined from the speed of the vehicle and its constant
parameters.
The same approach can be considered to estimate the emissions of other
pollutants.
2.2.2. Probabilistic methods

280 The previous models estimate emissions and energy consumption as a func-
tion of real vehicle operation variables (e.g. speed and acceleration, power de-
mand, engine mechanics, etc.).
However, these data are not always available. One may obtain the velocity
through microscopic traffic model simulation. But such models can be difficult
285 to implement, especially on a large spatial scale with a lack of precise knowl-
edge about the traffic situation, and can lead to unrealistically smooth velocity
profiles. Hence, probabilistic models, based on random velocity disturbances,
have been proposed in the literature.
The general idea of the random velocity disturbances approach is to run the
290 emission and energy consumption models while replacing, for a given route, the
actual speed of the vehicle by an approximate second-by-second speed profile
built from a deterministic and a stochastic component as
ṽ = v̄ + Θ (19)
where v̄ is the average traffic speed estimated from the road attributes provided
by a geographical information system (e.g. speed limit, traffic signs, road grade,
295 etc.) and Θ is a random variation in velocity for the subject vehicle.
One possibility is to consider a spatial distribution of speed or acceleration
based on driving cycles or statistical distributions (Burghout, 2004).
Karbowski et al. (2014) combine Markov chains with deterministic route
attributes to generate the speed profile. In this model, Θ is adjusted according
11
300 to
P (X(k + 1) = Xi |X(k) = Xj ) = MT P (i, j) (20)
where X(k) = [v(k) a(k)]T is the state vector of the vehicle at time step k, and
the transition probability matrix MT P is built from real data.
Another probabilistic model is the Motor Vehicle Emission Simulator (MOVES),
presented by Wu et al. (2014). The aim of this method is to make the veloc-
305 ity trajectory more realistic. Thus, it is assumed that vehicle detector stations
provide an estimation of v̄. The random variation in velocity is defined as
Θ(k) = ṽ(k − 1) + a(k − 1) − v̄(k) (21)
A procedure to determine the acceleration a is presented in Wu et al. (2014).

Probabilistic approaches are a solution in case of lack of information about
the vehicle dynamics. By construction, they are less accurate than models
310 based on the actual speed, but can be effectively used to estimate emissions and
energy consumption (Kubička et al., 2016). To improve these methods, traffic
prediction models could be integrated to determine v̄ (cf. Section 3.1.2).
Note that the variability of certain unobserved parameters between vehicles
(e.g. temperature, Reid vapor pressure) can affect the emissions and energy con-
315 sumption. These issues can be addressed by introducing probabilistic correction
factors (Frey and Zheng, 2002).
3. Emission and energy consumption models for traffic vehicular flows
The emission and energy consumption models presented in Section 2 are

microscopic. They estimate emissions and energy consumption based on the
320 instantaneous operating variables of individual vehicles, that can be obtained
through microscopic traffic models. But on a network scale, they have the known
disadvantages of high computational load, as their computation time increases
sharply with the number of vehicles. The instantaneous operating variables can
also be measured, but the data for so many vehicles are very difficult to obtain
325 and process.
For large scale control purposes it is necessary to develop macroscopic models
that use aggregate network or link-based data to estimate global emissions and
energy consumption. These models are more coarse but also simpler to use and
allow for faster computation. They are based on the traffic variables presented
330 in Table 2.
In this section, we first review how to determine the traffic kinematics, then
we present different emission and energy consumption models that can be set
up.
3.1. Traffic kinematics

335 To determine the traffic kinematics, it is possible to measure the average
speed of vehicles, or to use a traffic model based on fluid dynamics.
12
i Cell index [-]
k Discrete Time index [-]
L Length of link [m]
Ni (k) Number of vehicles in cell i at time step k [veh]
v Traffic speed [m.s−1 ]
v̄ Average traffic speed [m.s−1 ]
vmax Maximum speed, forward wave speed [m.s−1 ]
δ Discrete-time step size [s]
δL Discrete-space cell length [m]
ρ Vehicle density [veh.m−1 ]
ρcr Critical density, i.e. associated with the maximum flow [veh.m−1 ]
ρM Maximum possible vehicle density [veh.m−1 ]
ϕ Traffic flow [veh.s−1 ]
ϕi (k) Traffic flow entering cell i during
time interval [kδ, (k + 1)δ] [veh.s−1 ]
ϕM Maximum possible traffic flow [veh.s−1 ]
w Backward wave speed [m.s−1 ]
Table 2: Traffic variables and indices used in emission and energy consumption models.
3.1.1. Average speed

The average speed of the traffic on each link i is defined as
T N (i,k)
1 X 1 X
v̄(i) = v j (k) (22)
T N (i, k) j=1
k=1
where T is the number of time steps on which the average speed is performed,
340 N (i, k) is the number of vehicles on link i at time step k and v j (k) denotes the
speed of vehicle j at time step k. In the following, v̄(i) is referred to as v̄ for
simplicity.
The average speed can be provided using fixed sensors or Floating Car Data
(FCD) methods, like the smartphone devices of the drivers for example. Sim-
345 ilarly, the number of vehicles N (i, k) can be provided by induction loops or
cameras.
3.1.2. Fluid-based models

The traffic kinematics can also be determined through dynamic fluid-based
traffic models that describe the evolution of the traffic in the network as a fluid
350 in a pipe. Some overviews presenting this kind of model can be found in Ferrara
et al. (2018a,c); van Wageningen-Kessels et al. (2015); Hoogendoorn and Bovy
(2001).
This approach provides the traffic variables, i.e. ρ(x, t), v(x, t), and ϕ(x, t),
at given position x and time t. It considers the traffic speed as a function of the
355 position x and the time t. Therefore, unlike the average speed approach, these
13
models reflect the speed differences along links and provide a dynamic traffic
speed.
Some macroscopic traffic models are reviewed in the following. They are all
based on the following conservation law
∂ ∂
ρ(x, t) + (ρ(x, t)v(x, t)) = 0 (23)
∂t ∂x
360 Some of these models are continuous and others are spatially and temporally
discretized. A distinction is made between first and higher order models.
– First order models
• Lighthill-Whitham-Richard model
Lighthill and Whitham (1955) and Richards (1956) assume that v depends
365 only on ρ. Hence, the flow can be expressed as a function of only ρ as
ϕ = ρv(ρ) = Φ(ρ) (24)
The conservation law presented in (23) can then be expressed as
∂ ∂
ρ+ Φ(ρ) = 0 (25)
∂t ∂x
where Φ is a strictly concave C 1 function defined on [0, ρM ] and satisfying Φ(0) =
Φ(ρM ) = 0.
The relationship ϕ = Φ(ρ) is called the fundamental diagram. The most
370 common fundamental diagrams are listed in Table 3 (Garavello et al., 2016), in
which v0 is a positive constant.
p
Greenshields et al. (1935) Φ(ρ) = ρvmax 1 − ρρM , p∈N
M
ρ
Greenberg (1959) Φ(ρ) = ρv0 ln ρ

Underwood Φ(ρ) = ρvmax exp − ρρM

California Φ(ρ) = ρv0 ρ1 − ρ1M
M
Trapezoidal (Daganzo, 1994) Φ(ρ) = min ρv max , ϕ , M (ρM − ρ)w
Triangular (Newell, 1993) Φ(ρ) = min ρvmax , (ρ − ρ)w
Table 3: List of most common fundamental diagrams
• Cell transmission model

Daganzo (1994) proposes the cell transmission model (CTM) which is a
temporally and spatially discretized version of the LWR model based on the
375 triangular or the trapezoidal fundamental diagram. The model is defined as
 ρi (k + 1) = ρi (k) + δδL (ϕi (k) − ϕi+1 (k))


(26)
ϕi (k) = min ρi−1 (k)vmax , ϕM , w(ρM − ρi (k))

14
• Variable-length model
In order to depict density evolution and track the congestion front, Canudas-
de-Wit (2011) proposes the variable-length model (VLM) for highway traffic
modeling. Illustrations are given on a closed ring road and on an urban road
380 with traffic lights in Canudas-de-Wit and Ferrara (2018).
The VLM is also a discrete version of the LWR model based on the triangular
fundamental diagram. The idea is to model any road section with only two
lumped cells that are variable in length: an upstream cell in free flow and
a downstream congested cell. Consider a road section of length L, then the
385 length of the free and the congested cells will respectively be L − l and l.
The main advantage of the VLM is that it is based on only three state vari-
ables: density in the upstream free cell ρf , density in the downstream congested
cell ρc , and position of the congestion front l. The model reads
1
ρ̇f = [ϕin − ϕ(ρf )] L−l






ρ̇c = [ϕ(ρc ) − ϕout ] 1l (27)



 ˙ ϕ(ρf )−ϕ(ρc )

l= ρc −ρf
where the interface flows ϕ(ρf ) and ϕ(ρc ), which correspond to the demand of
390 the free cell and the supply of the congested cell respectively, can be expressed
as
ϕ(ρf ) =ρf vmax
(28)
ϕ(ρc ) =w(ρM − ρc )
ϕin and ϕout are the inflow and outflow at the boundaries of the section of
length L. They are defined as
ϕin = min {Din , Sf }
(29)
ϕout = min {Dc , Sout }
where Din and Sout are respectively the input demand and the output supply.
395 Dc and Sf are
Dc = min {ρc vmax , vmax ρcr (vmax )}
(30)
Sf = min w(ρM − ρf ), vmax ρcr (vmax )

where ρcr (vmax ) is the critical density relative to vmax . It is defined as

wρM
ρcr (vmax ) = (31)
vmax + w
De Nunzio et al. (2014) propose to adapt the VLM to the urban environment
by considering a binary variable ζ multiplying the boundary flows in (29) to
model the behavior of traffic lights, as

1 , if the traffic light is green
ζ= (32)
0 , else
15
400 • Link transmission model
Yperman (2007) proposes the link transmission model (LTM), which is a
discrete version of the LWR model based on the triangular fundamental diagram,
with only one cell per road. Therefore, computation times are reduced.
tot
The LTM introduces the notion of cumulative vehicle counts. Nup (kδ) and
tot
405 Ndown (kδ) are respectively the cumulative entering and exiting vehicle count of
a given link at kδ, based on given split ratios at intersections.
The maximum number of vehicles that can be sent by this link to the next
one during time interval [kδ, (k + 1)δ] is

tot L tot
Sboundary (k) = Nup (k + 1)δ − − Ndown (kδ) (33)
vmax
The maximum number of vehicles that can leave the considered link during
410 the time interval [kδ, (k + 1)δ] is
Slink (k) = ρM Lδ (34)
The number of vehicles sent by the link to the next one is then simply
S(k) = min {Sboundary (k), Slink (k)} (35)
In the same way, the number of vehicles R(k) received by the link is expressed
as 

 R(k) = min {Rboundary (k), Rlink (k)}



L
tot

Rboundary (k) = Ndown (k + 1)δ − w + ρM L − Nup
tot
(kδ) (36)




Rlink (k) = ρM Lδ

• Queue models
415 Queue models are interested in the length of the queues at the end of each
link i. An example is the Berg-Lin-Xi (BLX) model, presented by Van den Berg
et al. (2007) and Lin and Xi (2008). Lin et al. (2012) propose an extension of
the BLX model. Like the LTM, the BLX model considers flows between the
links.
420 The queue on link i is composed of Niq vehicles. When the traffic light is
green, the number of vehicles entering cell i from the upstream cell during time
interval [kδ, (k + 1)δ] is
q
(k) + δϕi−1 (k), Si (k), δϕM

δϕi (k) = max 0, min Ni−1 (37)
where Si (k) denotes the available storage of link i at time step k, expressed in
number of vehicles.
425 The queue length and the available storage can be expressed as
 q q
 Ni (k + 1) = Ni (k) + δ(ϕi (k) − ϕi+1 (k))
(38)
Si (k + 1) = Si (k) + δ(ϕi+1 (k) − ϕi (k))

16
The total number of vehicles in link i at time step k can be determined as
Ni (k) = ρM L − Si (k) (39)
• Summary
The LWR model and its discrete variations, presented above, are simple first
order mathematical representations of the traffic inspired by fluid mechanics.
430 They are based on fundamental diagrams that associate ϕ to ρ. These models
are able to capture realistic traffic phenomena such as shock waves, physical
queues and queue spillbacks (Garavello et al., 2016).
However, first-order models based on the fundamental diagram are not suf-
ficient to capture unstable traffic variations caused by the inertia of vehicles
435 because they assume that v is always in equilibrium. Consequently, they have
limitations in capturing complex traffic phenomena such as stop-and-go waves,
capacity drops and phantom jams (formation of clusters of cars with high densi-
ties due to the driving style of road users Kerner and Konhäuser (1993)). These
must be taken into account in order to best estimate emissions and energy con-
440 sumption.
– Second order models
Second order models have been developed in order to capture more realistic
traffic behavior in congested areas. They still consider the equation for the
conservation of vehicles presented in (23) and use the fundamental diagram to
445 determine the steady state of the system, but they have an additional equation
for the conservation of momentum.
• Payne-Whitham model
An example of a well known second order model is proposed in Payne (1971).
The model has the following form

∂t ρ + ∂x (ρv) = 0
(40)
∂t v + v∂x v + ρ1 ∂x (p(ρ)) = τ1 (ve (ρ) − v)
450 where ve (ρ) is the equilibrium speed given by the fundamental diagram, and
p(ρ) is analogous to the pressure in the fluid dynamics equations and depends
on the density (Piccoli and Tosin, 2009).
The anticipation term ρ1 ∂x (p(ρ)) models the reaction of vehicles, i.e. accel-
eration or deceleration, to the variations of ρ. The relaxation term τ1 (ve (ρ) − v)
455 models the tendency of vehicles to travel from v towards ve (ρ) within a time
τ > 0 that represents the time needed by the vehicles to adjust their actual
speed to ve (ρ).
The second equation of (40) is the acceleration equation. Whitham (1974)
proposes to simplify the model by considering p(ρ) as a constant. Other expres-
460 sions for this term exist, they are presented in Garavello et al. (2016), as well
as the modeling of an additional viscous term in this equation.
17
• Aw-Rascle-Zhang model
Daganzo (1995) highlights some limitations of the Payne-Whitham model
presented above. In particular, the model allows the vehicles to travel with
465 negative speed.
To tackle this problem, Aw and Rascle (2000) and Zhang (2002) propose the
following model
∂t ρ + ∂x (ρv) = 0
(41)
∂t (v + p(ρ)) + v∂x (v + p(ρ)) = 0
where the pressure term p may be defined as
p(ρ) = ργ , γ>0 (42)
• METANET model
470 Messmer and Papageorgiou (1990) present the METANET model which is
a discrete version of the Payne-Whitham model presented in (40). It reads
 δ
 ρi (k + 1) = ρi (k) + δL (ϕi (k) − ϕi+1 (k))





vi (k + 1) = vi (k) + γδ1 [ve (ρi (k)) − vi (k)] (43)




+ δδL vi (k)[vi−1 (k) − vi (k)] − γ2γδ[ρ i+1 (k)−ρi (k)]


1 δL [ρi (k)+γ3 ]
and the authors propose the following fundamental diagram to define the equi-
librium speed ve (ρ)
γ
1 ρi (k) 4

ve (ρi ) = vmax exp − (44)
γ4 ρcr
where γ1 to γ4 are model coefficients.
475 METANET was originally introduced to capture traffic phenomena on high-
ways. The proceeding of flows between the segments is fully presented in Mess-
mer and Papageorgiou (1990).
– Phase transition and higher order models
Second order models generally have higher computation times. Phase tran-
480 sition models are a good alternative to the extent that they behave like the
classic LWR model when the traffic is free and like a second-order model when
the traffic is congested. This allows to capture complex traffic phenomena while
keeping reasonable computation times for free-flow traffic.
Colombo (2002) proposes the following phase transition model
485 • For free flow traffic, the author considers the LWR model, presented in
(25), with the Greenshields fundamental diagram (cf. Table 3, with p = 1).
• For congested traffic, v cannot be considered as a function only of the
density anymore. In this case, the density-flow points are scattered in a
two-dimensional region.
18
490 The associated hybrid fundamental diagram is shown in Fig. 4. The following
second-order model is then introduced for high densities

∂t ρ + ∂x (ρv) = 0
(45)
∂t q + ∂x ((q − Q)v) = 0
where q is the momentum, Q is a parameter of the road considered, and v is

expressed as
ρ q
v = 1− M (46)
ρ ρ
Figure 4: Fundamental diagram of the phase transition model, representing the free region
Ωf and the congested region Ωc (Colombo, 2002) [Copyright c 2002 Society for Industrial
and Applied Mathematics. Reprinted with permission. All rights reserved].
Finally, higher-order models exist but they are less appropriate for emissions
495 and energy consumption estimation as their computation times are higher. An
example of third order models, where the additional equation is for the variance
of the speed, can be found in Helbing (1995). This method is useful to describe
the increase of the speed variance just before traffic jams occur.
– Network-wide extension
500 The traffic models presented above can be extended across a network. The
junctions represent a very important part of the extended model. Basically, each
junction can be reduced to a combination of simple merge and diverge junctions
(Garavello et al., 2016). A complete overview of macroscopic node models can
be found in Tampère et al. (2011). The authors present macroscopic node model
505 instances both for signalized and unsignalized intersections.
In the case of the extended CTM, a fundamental diagram is associated with
each link, each link being partitioned into uniform cells. An urban version
of the CTM is proposed in Xie et al. (2013). The authors use turning ratios
assigned to intersections and distinguish two possibilities. First, a cell preceding
19
510 an intersection can be composed of one traffic light. Such cells have one unique
queue, and all the vehicles merge into it. Second, the cell can be divided into
sub cells so that each direction has its own traffic light.
Similarly, the LTM can be extended considering the flows sent and received
by links (Garavello et al., 2016). Regarding queue models, Lin et al. (2012)
515 consider the case of links with multiple junctions (connected to several upstream
and downstream links) and, for control purposes, present the S model, which is
basically a simplification of the BLX model, with a time interval equal to the
traffic-light cycle.
The network-wide extension approach is similar for second order traffic mod-
520 els. For example, Garavello et al. (2016) present the extension of the Aw-Rascle-
Zhang model on a network scale. A more detailed description of this model at
junctions can be found in Herty and Rascle (2006).
For control purposes, De Nunzio et al. (2014) suggest to simplify the VLM
by assuming an average continuous flow through the traffic lights by replacing
T
525 the binary variable ζ with Tgreen cycle
, where Tgreen and Tcycle denote respectively
the green phase time and the cycle time of the traffic light. This method is
inspired by store-and-forward models, originally suggested by Gazis and Potts
(1963). It allows to describe the urban traffic without using binary variables.
Hence, polynomial complexity control methods can be applied to the system,
530 which allows for consideration of large-scale networks. However, due to this sim-
plification, the effect of offsets between traffic lights of successive intersections is
not depicted. Moreover, the oscillations of the system (stop-and-go waves, prop-
agation waves, etc.) are not represented, which is a crucial point for emissions
and energy consumption estimation (Hall, 2012; Aboudolas et al., 2009).
535 3.2. Emission and energy consumption meta-models

In Section 3.1, we reviewed some methods to determine the traffic kinematics,
either by measuring the static average speed (cf. Section 3.1.1), or by using
dynamic fluid-based traffic models (cf. Section 3.1.2). In this section, we present
the meta-models used to calculate emissions and energy consumption from the
540 traffic dynamics, considering either approach.
3.2.1. Meta-model associated with static average speed-based approaches

Emissions and energy consumption can be calculated by considering an av-
erage speed-based approach. This is done by a meta-model whose general pro-
cedure is illustrated in Fig. 5, and operation steps are presented below.
545 1. The average speed v̄ and the number of vehicles N are measured, or esti-
mated.
2. The emission or energy consumption rate of a single vehicle Jy is calculated
from v̄ using a microscopic emission and energy consumption model (cf.
Section 2).
550 3. Jy (v̄) is then multiplied by N to approximate the total emission or energy
consumption rate.
20
v̄ Microscopic Jy (v̄) Macroscopic
Measurements emission/energy model emissions/energy
Figure 5: Structure of the emission and energy consumption meta-model associated with static
average speed-based approaches.
Note that this procedure can be conducted at different scales. The average
speed v̄ and the number of vehicles associated N can refer to a single link of
a network, if the data are available, or to a larger spatial area. Moreover, the
555 duration between two successive measurements usually depend on the measuring
devices. These issues are addressed in Section 3.3.
This meta-model can be associated either with a data-based or a physical
microscopic emission and energy consumption model. These approaches are
detailed below. Note that they involve measuring, or estimating, the number of
560 vehicles on the roads under consideration.
Some authors propose to associate the meta-model with a data-based mi-
croscopic emission and energy consumption model.
For instance, Boriboonsomsin et al. (2012) propose the following regression-
565 based model in order to estimate the fuel use rate of a single vehicle

ln Jfspat 2 3 4
uel = β0 + β1 v̄ + β2 v̄ + β3 v̄ + β4 v̄ + β5 α (47)
where β0 to β5 are the regression coefficients.

Another common approach to estimate emissions and energy consumption
on a large spatial scale is to associate this meta-model with a microscopic model
based on aggregated data-driven emission or energy consumption factors Jy (v̄, θ)
570 that depend on the traffic average speed v̄ and some vehicle parameters θ.
Let Ω be the set of possible parameters sets. Aggregated factors are usu-
ally simply the mean values of experimental measurements and are typically
expressed in mass of pollutant emitted (or mass of fuel consumed) per vehicle
and per unit distance traveled. Hence, the total emission or energy consump-
575 tion rate (per distance traveled), i.e. the output of the meta-model, of a link i
containing N (i, θ) vehicles with the set of parameters θ is given by
X
Jyi = N (l, θ)Jy (v̄, θ) (48)
θ∈Ω
In practice, detailed information on the fleet composition is not available.

Hence, a reference set can be considered, i.e. all the vehicles have the same
parameters θ̄, and the emission or energy consumption rate on link i simply
580 becomes:
Jyi = N (i, θ̄)Jy (v̄, θ̄) (49)
21
The COPERT (COmputer Programme to calculate Emissions from Road
Transport) model (Ntziachristos et al., 2009) developed by the European En-
vironment Agency is based on this method. Several vehicle parameters are
included in θ: the vehicle type (passenger car, light commercial vehicle, heavy
585 duty vehicle, L-category vehicle), the fuel type, the engine displacement and its
registration date. The sets of parameters of all the vehicles constitute the vehi-
cle fleet composition. An example of emission factors obtained with COPERT
for different types of vehicle according to the speed is given in Fig. 6.
Figure 6: Fuel consumption factors of different gasoline passenger cars calculated with COP-
ERT (Sobrino et al., 2014) [Published with permission of Networks and Spatial Economics].
Hausberger (2009) proposes the HBEFA (HandBook Emission FActors for

590 road transport) model, which is more precise. This method additionally con-
siders the driving conditions (highways, urban roads, stop-and-go traffic) and
the volume-to-capacity ratio (number of vehicles divided by the capacity of the
link), which is a dynamic variable, to determine Jy (v̄, θ).
The accuracy can also be improved by multiplying the emission and energy
595 consumption factor Jy (v̄, θ) by a congestion correction factor, as does the TEE
(Traffic Energy and Emissions) model. The objective is to represent the effect
of congestion on emissions and energy consumption. The congestion correction
factor depends on the average speed, the traffic-light timing, the link length, and
the traffic density (Negrenti, 1999). These variables and parameters are used to
600 estimate the time spent in each traffic situation (cruising, acceleration, decel-
eration and idling) and thus reflect the speed variability along the considered
road link. The corresponding speed profile can then be reconstructed.
One limitation of these aggregated factors models is that the emission and
energy consumption factors are not fundamental, as they depend on the driving
22
605 cycle used during the measurements.
• Physical model
It is also possible to use this meta-model by associating it with a physical
emission and energy consumption model.
For example, Jurik et al. (2014) propose to use the following microscopic
610 physical model to estimate the energy consumption of a vehicle on link i
r
i E (i) + (ν − 1)E p (i) , if E p (i) ≤ 0
Ey (i) = Jy L(i) = (50)
E r (i) , if E p (i) > 0
where L(i) is the length of the link i, and ν ∈ [0, 1] is the downhill potential
energy recuperation coefficient. The resistance and the potential energies are
respectively given by
d
E r (i) = ACd v̄ 2 L(i) + M gCr L(i) cos α
2 (51)
E p (i) = M gL(i) sin α
To model more precisely the speed change at an intersection, De Nunzio

615 et al. (2017) introduce a transition speed at the interface between two links of
respective average speeds v̄bef ore and v̄af ter defined as
v̄bef ore + v̄af ter
vtransition = β (52)
2
where β ∈ [0, 1] is a parameter depending on the type of interface (e.g. stop
sign, traffic light, turning movement, etc.). This transition speed can be intro-
duced to any model similar to the one presented in (50) to (51) to better model
620 intersections.
3.2.2. Meta-model associated with dynamic fluid-based models

Emissions and energy consumption can be calculated by considering the
fluid-based models dynamics. This is done by another meta-model whose general
procedure is illustrated in Fig. 7, and operation steps are presented below.
625 1. First, a dynamic fluid-based traffic model is chosen (cf. Section 3.1.2). It
provides the traffic variables, i.e. ρ(x, t), v(x, t), ϕ(x, t).
2. Then, these variables are processed by an interface to generate groups of
vehicles g(x, t) sharing the same speed and acceleration. The interface
calculates the speed, acceleration and number of vehicles of each group.
630 They are respectively denoted v(g(x, t)), a(g(x, t)) and N (g(x, t)).
3. A microscopic emission and energy consumption model is chosen (cf. Sec-
tion 2). It provides the emission or energy consumption rate Jy (v, a) of
a vehicle of group g(x, t) using the outputs v(g(x, t)) and a(g(x, t)) of the
interface.
23
ρ(x, t)
v(x, t) v(g(x, t))
ϕ(x, t) a(g(x, t)) Microscopic
Macroscopic
Interface emission/energy model
traffic model
N (g(x, t)) Jy (v, a)
Macroscopic emissions/energy
Figure 7: Structure of the emission and energy consumption meta-model associated with
dynamic fluid-based traffic models.
635 4. The emission or energy consumption rate Jy (v, a) of a vehicle of group

g(x, t) is multiplied by the number of vehicles in the corresponding group
N (g(x, t)) to provide the total emission or energy consumption rate of
group g(x, t).
The procedure presented above is repeated as many times as there are
640 groups. It is important to note that this generation of groups with homoge-
neous operation variables relies on the spatial and temporal discretizations of
the traffic model. This issue is addressed in Section 3.3.
The procedure of the interface presented above is detailed in the following
(Zegeye et al., 2013).
645 To compute emissions and energy consumption, the employed traffic models
are often discrete both in time and in space. Hence, two acceleration components
have to be considered: the temporal and the spatial accelerations:
- The temporal acceleration describes the change in speed of vehicles within
a cell from one time step to the next. It only applies to the vehicles that
650 remain in the cell. It is expressed as
vi (k + 1) − vi (k)
atemp
i (k) = (53)
δ
The number of vehicles subject to this acceleration, i.e. that stay within
the cell i from time step k to time step k + 1 is equal to
Nitemp (k) = δL ρi (k) − ϕout
i (k)δ (54)
where ϕout
i (k) is the outflow of cell i during time interval [kδ, (k + 1)δ].
The first term represents the number of vehicles initially in cell i at time
655 step k.
These vehicles constitute a group g(x, t) mentioned above. There are as
many groups of this type as there are cells in the spatial discretization of
the traffic model.
24
- The spatial acceleration describes the change in speed of vehicles moving
660 from cell i to cell j. It is defined as
vj (k + 1) − vi (k)
aspat
i,j (k) = (55)
δ
The number of vehicles subject to this acceleration, i.e. that move from
the cell i to cell j during time interval [kδ, (k + 1)δ] is
spat
Ni,j (k) = ϕi,j (k)δ (56)
where ϕi,j (k) is the flow of vehicles moving from cell i to cell j.
These vehicles constitute a group g(x, t) mentioned above. There are
665 as many groups of this type as there are cells interfaces in the spatial
discretization of the traffic model.
Ultimately, a generic formulation to calculate the emissions and energy con-
sumption rate in a network made of n cells is
Pn
Jynetwork (k) = Jy (atemp
i (k), vi (k))Nitemp (k)+
Pi=1
n P n spat spat (57)
i=1 j=1 αi,j Jy (ai,j (k), vi (k))Ni,j (k)
where αi,j is a binary variable equal to 1 if cells i and j are connected in the sense
670 that vehicles can move directly from cell i to cell j, αi,j equals zero otherwise.
The first term of (57) refers to the emissions and energy consumption of vehicles
staying in the same cell from time step k to k + 1, and the second term refers
to those of vehicles moving from one cell to another.
To estimate emissions and energy consumption more precisely, this calcula-
675 tion can be done by differentiating classes of vehicles. In that case, the function
Jy can consider the real parameters of the vehicles instead of average values.
This meta-model can be associated either with a data-based or a physi-
cal microscopic emission and energy consumption model. Some examples are
presented below. Naturally, the meta-model procedure is generic and can be
680 adopted to other models.
Some authors propose to use this meta-model by associating it with a data-
based emission and energy consumption model. For example, Zegeye et al.
(2013) propose to integrate the macroscopic traffic second order model METANET
685 with the microscopic data-driven emission and fuel consumption model VT-
micro. The resulting meta-model, called VT-macro, is mainly suitable for mod-
eling emissions and energy consumption on highways.
Similarly, Lin et al. (2013) suggest to associate the traffic first order S model
with VT-micro in an urban network. The authors present a set of possible
690 behaviors for the vehicles (e.g. free, idling, accelerating, decelerating, start-
and-stop behavior). Another use of the meta-model in an urban environment
can be found in Jamshidnejad et al. (2017), in which the authors propose the
same models association.
25
• Physical model
695 It is also possible to use this meta-model by associating it with a physical
emission and energy consumption model. For example, De Nunzio et al. (2014)
develop a method based on the VLM and a physical approach to determine
energy consumption. This model considers only the spatial component of accel-
eration. In each cell, energy consumption is determined from the cell average
700 speed (either free or congested), considering zero temporal acceleration. At the
interface of the cells, the energy is calculated based on the following spatial
acceleration
vj − v i
aspat
i,j = min amax , (58)
δ
where the maximum acceleration amax is a model parameter. Note that time
does not appear in the formulation of De Nunzio et al. (2014) because the
705 analysis is performed at steady state.
3.3. Spatial and temporal discretizations

The spatial and temporal discretizations of the methods used are a crucial
point in emissions and energy consumption estimation: a compromise has to be
found between precision and computation time.
710 • Spatial discretization
Concerning the use of the average speed meta-model, measurements of aver-
age speed and number of vehicles made on a road level would naturally give bet-
ter results than measurements made on a larger spatial scale. But this depends
mainly on the devices used to monitor the traffic. Some average speed-based
715 meta-models consider a fine spatial discretization in order to be compatible with
urban networks (e.g. COPERT Street Level (Rai et al., 2017)).
When using the meta-model associated with dynamic fluid-based models,
the choice of the spatial discretization step size should be given some thought.
A balance concerning the number of cells and their length has to be found in
720 order to satisfy the desired accuracy without excessively increasing computation
times.
• Temporal discretization
Some authors have proposed methods to use the average speed meta-model
with high-frequency data inputs, i.e. average speeds and number of vehicles
725 updated at high frequency. For example, Lejri et al. (2018) propose a method
to adapt the COPERT emission and fuel consumption model to high-frequency
data inputs. This kind of approach is more precise. However, it is essential
to note that the average speed-based meta-model is static. In other words,
even with high frequency data inputs, emissions and energy consumption are
730 calculated for successive average speeds, but do not consider the acceleration
of vehicles, yet crucial to fully characterize emissions and energy consumption
(Ahn and Rakha, 2008).
26
The dynamic fluid-based meta-model considers vehicles accelerations. There-
fore, this approach is more precise, but its crucial point is the estimation of the
735 acceleration. In the formulation proposed in Section 3.2.2, the choice of δ must
lead to realistic accelerations values while respecting the Courant–Friedrichs–Lewy
(CFL) condition expressed as
2δvmax ≤ δL (59)
• Summary
In other words, adopting a dynamic fluid-based approach with very long
740 time step size and length of cells is similar to having an average speed-based
approach. The difference would be that the data are obtained by simulation
instead of being measured.
Finally, the most precise way to calculate emissions and energy consump-
tion at a large spatial scale would be to use a microscopic traffic model and
745 to associate it with a microscopic emission and energy consumption model (cf.
Section 2). In fact, this approach is the only one able to reflect differences in
microscopic drivers’ behavior (e.g. sudden deceleration, merging, lane chang-
ing). When traffic is congested, these can result in shock waves causing traffic
breakdown, that a macroscopic traffic model cannot depict (Khondaker and
750 Kattan, 2015b). However, this method is not possible at large scale because
of the enormous computation times generated by the large number of vehicles
considered. Schiper (2017) proposes a statistical approach to process this large
amount of data by introducing sampling methods. The author suggests to esti-
mate emissions and energy consumption only in some relevant locations of the
755 network, and to extend the estimations at larger scales.
4. Single vehicles control design for emission and energy consumption

reduction
In Sections 2 and 3, emission and energy consumption models have been pre-
sented for single vehicles and for traffic flows. In this section, we review some
760 control strategies for single vehicles aiming at limiting emissions and energy
consumption. They can be mostly categorized into eco-driving, i.e. computing
a vehicle speed trajectory that minimizes the emissions or energy consumption
along a given route, and eco-routing, i.e. planning a minimum energy or emis-
sions route. An excellent overview of the existing vehicle control strategies is
765 given by Guanetti et al. (2018).
4.1. Eco-driving
Eco-driving consists in computing a vehicle trajectory that minimizes the
emissions or energy consumption along a given route, under technical (speed,
acceleration and brake characteristics of the vehicle) and environment (traffic,
770 traffic signs, traffic lights, etc.) constraints (Guanetti et al., 2018).
27
We define the state vector of a vehicle at time step k as x(k) = [s(k), v(k)]T ,
where s and v respectively denote the vehicle’s position along the route and the
speed of the vehicle. Let Ftrac and Fb be respectively the traction force at the
wheels and the mechanical brake force, the objective of the eco-driving algorithm
775 is to find at each time step k the input vector u(k) = [Ftrac (k), Fb (k)]T that
minimizes the emissions or the energy consumption calculated by the function
g. The objective function g is similar to those presented in Section 2. It may
consider vehicle parameters like its mass and parameters of the environment like
the road slope, usually provided by a Geographic Information System (GIS).
780 Sciarretta et al. (2015) formulate the eco-driving optimization problem as
follows
n−1
X
minimize g(x(k), u(k)) (60)
u0 ,...,un−1
k=0
subject to

x(k + 1) = f (x(k), u(k)), 

0 ≤ s(k) ≤ sf ,



vmin (k, s(k)) ≤ v(k) ≤ vmax (k, s(k)), ∀k ∈ [0 .. n − 1]
Ftrac,min (v(k)) ≤ Ftrac (k) ≤ Ftrac,max (v(k)), (61)




Fb,min ≤ Fb (k) ≤ Fb,max ,

x(0) = x0 ,
x(n) = xf .
The state of the vehicle at time step k + 1 is given by the following function
based on the vehicle dynamics

s(k) + δv(k)
f (x(k), u(k)) = δ (62)
v(k) + M (Ftrac (k) − Fb (k) − Fres (k))
785 where Fres is the resistance force, expressed in (8). The technical limits of the
vehicle consist in bounding the input variables Ftrac and Fb as indicated in
(61). The function vmin and vmax define a convex constraint set that takes into
account the environment constraints associated with speed limits, traffic lights,
traffic signs, road curvature, etc. x0 = [s0 , v0 ]T and xf = [sf , vf ]T are the
790 initial and final constraints of the eco-driving problem.
The eco-driving problem considering Ftrac and Fb as control inputs is per-
fectly compatible with autonomous vehicles, which include the control in the
longitudinal and lateral directions, as it gives instructions to the powertrain.
However, it is expected from an eco-driving problem for human drivers to re-
795 turn an advisory speed profile the user can follow. In that case, the algorithm
may return at each time step k the speed instruction v(k + 1) calculated with
(62) instead of Ftrac (k) and Fb (k). Another solution is to directly formulate
the optimization problem considering the recommended maximal speed of the
vehicle as the control input (Ozatay et al., 2014a; Boehme et al., 2013).
800 Sciarretta et al. (2015) present several algorithms aiming at solving the eco-
driving problem given in (60) to (62). These solutions can either be offline, i.e.
28
consider all road characteristics known in advance, or online, i.e. make use of
real-time estimations on a vehicle immersed in its environment.
Many solutions can be used for offline optimization: dynamic programming
805 (Dib et al., 2012), Pontryagin’s minimum principle (Sciarretta et al., 2015) or
calculating the analytical solution (Ozatay et al., 2014b).
Online solutions allow to acquire more information in real time about the
upcoming route. For example, Hellström et al. (2009) propose a method with an
on board optimizing controller taking into account the road slope. In the case
810 of connected vehicles, one may also imagine a control design taking into account
the prediction of the upcoming traffic conditions and accordingly updating the
vmin and vmax constraints of (61). The main limitation of these online solutions
is the computation time as they are expected to be compatible with real-time
execution.
815 In an urban environment, eco-driving is complex because of the uncertainty
of traffic. In particular, it is very difficult to know the traffic-light cycles in ad-
vance as some signalized intersections have a variable phase duration depending
on the traffic level. Intelligent transportation systems and traffic infrastruc-
ture connectivity are expected to reduce this uncertainty (Dimitrakopoulos and
820 Demestichas, 2010).
If the traffic-light cycles are unknown by the eco-driving algorithm, Ozatay
et al. (2014a) propose a method that considers traffic lights as stop signs in the
optimization problem. Naturally, the driver is free not to follow the advised
velocity given by the algorithm in the case of green at a traffic light.
825 To take into account the uncertainty about traffic-light cycles, Sun et al.
(2018) consider a stochastic cycle timing that adds to the red-light duration a
random variable. To generate more realistic signal timings, Mahler and Vahidi
(2012) introduce for each intersection a time-varying probability of green based
on measured data. In the optimization process, solutions that pass through
830 time intervals with high green probability are then naturally preferred.
In the case of known and deterministic traffic-light cycles, many algorithms
can be used to solve the eco-driving problem. For example, Miyatake et al.
(2011) present a method based on dynamic programming, De Nunzio et al.
(2016) use Dijkstra’s shortest path algorithm, HomChaudhuri et al. (2017) de-
835 velop a method with model predictive control, and Seredynski et al. (2013)
implement a genetic algorithm. The principle of these algorithms is to add a
constraint on the crossing time at intersections.
To improve the safety and avoid rear-end collisions, Zhang and Cassandras
(2018) propose a control strategy for vehicles crossing an urban signal-free in-
840 tersection. The principle is to generate acceleration profiles for the vehicles in
order to cross the merging zone in a limited time while minimizing the accel-
eration. This approach is adapted for autonomous vehicles, but the authors
consider a mixed traffic in their simulation (autonomous and human-piloted
vehicles) and analyze the impact of the proportion of autonomous vehicles on
845 their acceleration. Human-piloted vehicles are subject to priority rules.
Eco-driving algorithms need information about the traffic situation in order
to be accurate. These data can be provided either by sensors or by a macroscopic
29
traffic model. Many parameters of the problem, like pedestrians or drivers
decision making, remain uncertain and unpredictable.
850 Autonomous vehicles raise the issue of their safety, but they offer prospects in
terms of energy savings as they can accurately track the instructions generated
by the eco-driving algorithm (Han et al., 2018). Moreover, if the autonomous
vehicles communicate with each other, they can reduce their energy consumption
by coordinating and forming micro-platoons along the route, even if they have
855 different origins and destinations (Lelouvier et al., 2017).
4.2. Eco-routing
Eco-routing consists in planning an emission or energy-minimal route, given
an origin and a destination. The function that attributes to each link the energy
consumption (or the emissions) of a vehicle traveling along this link is denoted
860 g.
In the case of static eco-routing algorithms, the function g depends only on
the link under consideration. In the general case, the function g depends on the
time t as the traffic situation in the network evolves over time.
Ericsson et al. (2006) present an eco-routing algorithm that classifies the
865 roads of the network into 6 groups, depending on their GPS data. Based on the
same data, a fuel consumption factor is calculated for each group. Then, the
function g assigns to each link its energy consumption, using the fuel consump-
tion factor and the length of the link. The authors introduce peak and off-peak
hours to model the evolution of the traffic during the day. Similarly, Boriboon-
870 somsin et al. (2012) propose to consider not only historical GPS data, but also
real-time vehicle velocity trajectories to estimate the energy consumption of
each link, i.e. build the function g.
Usually, eco-routing algorithms only take into account the energetic cost of
links and not the vehicle behavior at intersections. However, this aspect is cru-
875 cial in energy consumption estimation. To model the energy consumption at
intersections, De Nunzio et al. (2017) introduce a transition speed at the inter-
face between two links, given in (52). Traffic lights at intersections have also to
be considered. For example, Sun and Liu (2015) propose an eco-routing algo-
rithm based on a signalized traffic network in which the authors use a Markov
880 decision process to model the traffic.
To determine the energy-optimal route, heuristic searches can be imple-
mented (Nannicini et al., 2012). Kluge et al. (2013) propose another approach
as the authors solve a time-dependent eco-routing problem by using an extension
of Dijkstra’s algorithm.
885 In order to limit the computation time, eco-routing algorithms can consider
a constraint on the maximum travel time or distance to reduce the set of possible
solutions. Another possibility is to implement multi-objective eco-routing that
minimizes not only the energy consumption but also the travel time and distance
traveled. In this case, the solution proposed is a Pareto optimal route (Bertsekas,
890 1995; De Nunzio et al., 2017).
30
5. Traffic flow control design for emission and energy consumption
reduction
In Section 4, we presented vehicle-based control designs aiming at reducing

the emissions and energy consumption of a single vehicle. In this section, we
895 review some road-based control strategies to reduce the environmental impact
on a large spatial scale and for a large number of vehicles. These strategies
consist in regulating vehicular flow by controlling speed limits, traffic-light duty
cycles or offsets, split ratios at intersections or bifurcations, or mobile actuators
(e.g. autonomous vehicles). In the following, the control strategies are classified
900 according to the employed actuator. For each actuator, we review first the
strategies adapted to highways, and then the strategies that can be set up for
an urban environment.
The objective of the following frameworks is to determine via an optimization
method, at each control time step, the control inputs that minimize the traffic
905 emissions and energy consumption. Note that some of the papers presented
in the following do not explicitly minimize emissions or energy consumption.
Instead, they tend to mitigate congestion and eliminate shock waves through
density homogenization, vehicles interdistance equalization, etc. These methods
are likely to indirectly reduce emissions and energy consumption as they reduce
910 the number of acceleration and deceleration (Barth and Boriboonsomsin, 2008).
However, it is important to make careful analyses about the assessment of the
effect of congestion reduction on emissions and energy consumption. In fact,
their relationship depends on many factors such as the speed of the traffic (Fiori
et al., 2018).
915 5.1. Speed limits control

A first approach to regulate the flow in order to reduce the emissions and
energy consumption is to control speed limits. This corresponds to imposing
variable location-dependent speed limits across the road network.
Many works present variable speed limits adapted to highways. Some of
920 them do not aim at explicitly reducing emissions and energy consumption (e.g.
SPECIALIST method that eliminates shock waves (Hegyi et al., 2008)). An
increasingly common approach is to use reinforcement learning methods to op-
timize speed limits. Walraven et al. (2016) propose to follow this approach to
minimize the amount of time vehicles spend on the highway under consideration.
925 Some other works are explicitly oriented on emissions and energy consump-
tion reduction. Generally, they implement a multi-objective optimization that
minimizes also the travel time so that unrealistic solutions like speed limits
equal to zero are avoided. For example, Zu et al. (2018) express the energy con-
sumption minimization on highways as a convex quadratic optimization prob-
930 lem whose objective function is derived from the average speed-based COPERT
model. The density is expressed as a function of the speed, considering the
Greenshields fundamental diagram (cf. Table 3). Another approach is proposed
by Zegeye et al. (2012). The authors proposes a control design applicable to
31
highways in which the speed limits are determined by Model Predictive Con-
935 trol (MPC). Because of the non-convex nature of the objective function, Zegeye
et al. (2012) uses a multi-start local sequential quadratic programming method
to determine the control inputs.
MPC offers opportunities for traffic control as it is compatible with the
uncertainties of the traffic models, and it can also handle non-linear and non-
940 convex optimization. However, computation times have to be reduced to make
MPC tractable for real-time operation, especially when the number of control
inputs is too large. Hence, in order to use MPC for macroscopic traffic con-
trol without significantly compromising the performance, Zegeye et al. (2012)
proposes to use a parameterized MPC, more specifically called Rolling Horizon
945 Parameterized (RHP) control.
In RHP control, the control inputs are parameterized according to some
time-profiles and the optimization focuses on the parameters. The number of
parameters to optimize is smaller than the number of control inputs and the
set of possible solutions is generally smaller. This results in faster computation
950 times but also a loss of performance. For computation time issues, RHP control
is more suitable for real-time application than conventional MPC, but it still
may be too slow, depending on the considered system, the parameterization,
and the control time step.
Note that some authors propose to use approaches based on microscopic
955 traffic models to control variable speed limits on freeways. For example, Khon-
daker and Kattan (2015a) present an MPC-based approach to maximize mobil-
ity, safety and environmental benefit.
A hybrid approach proposed in Van den Berg et al. (2007) aims at controlling
speed limits for mixed urban and highway networks. The authors present an
960 MPC framework that minimizes the TTS.
In an urban environment, some works do not explicitly minimize emissions
and energy consumption. For example, Tajali and Hajbabaie (2018) present
an MPC framework aiming at harmonizing the speed within the network and
maximizing the outflows.
965 Other works explicitly aim at reducing emissions and energy consumption.
Taylor (2000) presents an approach to evaluate the impact of various speed
limits on emissions, energy consumption, and traffic congestion, without seeking
to optimize speed limits. De Nunzio et al. (2014) propose a method to find the
optimal speed limit of a road section. The traffic model considered is the VLM
970 presented in Section 3.1.2, and the same notations are used here. The control
input is vmax and the objective function is the weighted sum of the energy
consumption, the total time spent T T S, the instantaneous travel time IT T and
the total travel distance T T D. These new metrics are defined as follows
L−l l
IT T (ρ) = + (63)
vmax vc
T T D(ρ) = Tcycle ρf vmax L + [w(ρM − ρc ) − vmax ρf ]l

(64)
32
T T S(ρ) = Tcycle (ρf L + (ρc − ρf )l) (65)
A method based on shock waves theory to control speed limits in an urban
975 area has been proposed by De Nunzio and Gutman (2017) to optimize energy
consumption and T T S.
Panis et al. (2006) present a methodology to analyze the environmental im-
pact of speed limits in an urban environment. The authors use the microscopic
traffic model DRACULA and a data-based emission and energy consumption
980 model. A case study is conducted in Ghentbrugge, a neighborhood of the city
of Ghent, Belgium. Similarly, Liu and Tate (2004) propose to study the effect
of speed limits in an urban network by implementing Intelligent Speed Adapta-
tion (ISA). This system suggests, or imposes, speed limits to the driver through
in-vehicle electronic devices. Note that ISA only informs road users of the speed
985 limits, but does not calculate it independently for each vehicle. In other words,
it is just a communication device. In this study, the authors consider the speed
limits as inputs of the simulation, i.e. they can vary with locations but are fixed
over the simulation time period, and are not optimized. One may also consider
dynamic speed limits based on an optimization framework. The authors use the
990 DRACULA traffic model. One of the main limitations of this kind of approach
based on microscopic traffic models is that a lot of data are involved. They are
usually very difficult to obtain, and they cause long computation times.
Note that machine learning methods can be used to control variable speed
limits. For example, Zhu and Ukkusuri (2014) present a Reinforcement Learning
995 (RL) approach aiming at optimizing the total network throughput, the delay
time, and the emissions. The authors propose a case study conducted on the
Sioux Falls network.
A general overview of the theoretical background and the main strategies of
variable speed limits strategies can be found in Khondaker and Kattan (2015b).
1000 5.2. Traffic lights control

Another solution to reduce emissions and energy consumption using road-
based control is traffic signal timing optimization. On freeways, this control
strategy, known as ramp metering, can be applied on on-ramps, and it consists
in regulating the traffic flow entering the highway.
1005 Many ramp metering strategies do not explicitly optimize emissions and
energy consumption, but they aim at reaching a desired density. That is the
case of ALINEA method, presented in Papageorgiou et al. (1991), which uses a
feedback law and the traffic density measured downstream from the merge area.
Similarly, Pisarski and Canudas-de-Wit (2016) present an approach to balance
1010 the vehicle density on the freeway by formulating the optimization problem as
a non-cooperative Nash game.
Some authors express the ramp metering control approach as an optimiza-
tion problem aiming at directly reducing emissions and energy consumption. For
example, Csikós et al. (2011) present a multi-objective optimization based on
1015 a constrained LQ (Linear-Quadratic) control, minimizing both T T S and traffic
33
emissions on freeways. Pasquale et al. (2015) formulate the ramp metering con-
trol problem as a multi-objective nonlinear constrained optimization problem
considering the same objective function. These metrics are calculated consid-
ering both the traffic in the on-ramp and in the mainstream. The emissions
1020 are calculated using an average speed-based model based on COPERT. The
nonlinear optimization problem is solved with a specific version of the feasible
direction algorithm: the derivative backpropagation method RPROP. A specific
feature of this work is that the authors consider two classes of vehicles (cars and
trucks) individually controlled as determined by the optimization process.
1025 In an urban environment, many traffic signal timing optimization strategies
have been developed to control traffic-light cycles. Most of them do not explic-
itly optimize emissions and energy consumption. Instead, they minimize the
congestion by improving the throughput and reducing the delay. Some of these
strategies are: SCOOT (Split, Cycle and Offset Optimisation Technique) (Hunt
1030 et al., 1981), SCATS (Sydney Coordinated Adaptive Traffic System) (Lowrie
et al., 1982), RHODES (Real-time Hierarchical Optimized Distributed Effec-
tive System) (Mirchandani and Head, 2001), TUC (Traffic-responsive Urban
Control) (Dinopoulou et al., 2006), max-pressure (Varaiya, 2013).
Grandinetti et al. (2018) formulate the signal timing control problem as
1035 a CTM-based real-time convex optimization whose objective function is the
weighted sum of T T D, the density balancing and a regularization term that
penalizes abrupt changes in the control dynamics. The density balancing term
aims at homogenizing the density over the network. The algorithm is split into
subproblems whose sizes are independent of the network size, thus allowing for
1040 scalability.
Some authors explicitly consider emissions and energy consumption reduc-
tion. For example, Han et al. (2016) express the signal timing optimization in
an urban environment as an LTM-based Mixed Integer Linear Program (MILP)
optimizing both the delays and the emissions. Emissions are calculated as a
1045 function of the density of the links, by calculating the spatial and the tempo-
ral accelerations defined in Section 3.2.2. Similarly, Osorio and Nanduri (2015)
propose a meta-model that considers the simulations of T T S and fuel consump-
tion as well as their analytical approximations to solve the urban signal timing
optimization using a simulation-based optimization algorithm.
1050 MPC can be implemented in a traffic light control framework. For example,
Lin et al. (2013) present a method adapted for urban traffic networks based on
MPC. The authors consider a dynamic fluid-based meta-model associating the
S model and VT-micro to characterize emissions (cf. Section 3.2.2). The ap-
proach aims at reducing both the congestion and the emissions as the objective
1055 function considers the weighted sum of T T S and total emissions. Jamshidnejad
et al. (2018) propose a similar approach based on a gradient-based optimiza-
tion approach. The authors consider an extension of the S traffic model. The
objective function considers the weighted sum of T T S, total emissions, and the
absolute difference of two temporally successive control inputs, in order to avoid
1060 abrupt variations.
A more precise approach can be found in Stevanovic et al. (2009). The
34
authors propose to simulate the traffic dynamics through a microscopic traffic
model, namely VISSIM, and to calculate the emissions by using the CMEM
emission and energy consumption model (cf. Section 2.2). A signal timings op-
1065 timization is then conducted using VISGAOST, an optimization program based
on the stochastic nature of genetic algorithms. Although the authors propose
a case study on a road network composed of two suburban arteries, an online
optimization based on this method is not possible because of lengthy calculation
times. But such methods can be implemented for an offline optimization.
1070 RL methods can also be implemented to control traffic lights in an urban
network. For example, Khamis and Gomaa (2012) present a framework that
considers the microscopic dynamics of vehicles. The authors propose to approx-
imate the energy consumption metric by the average number of vehicles stops,
assuming that this performance index can be directly related to ecological issues.
1075 5.3. Coordinated speed limits and traffic lights control

To improve the results of road-based control, it is possible to coordinate
actuators such as speed limits and signal timing control.
For freeways control, Hegyi et al. (2005) propose a method to optimize T T S,
without considering emissions and energy consumption. The authors develop a
1080 MPC framework, in which the control inputs are speed limits and ramp meter-
ing.
Other authors have used coordinated speed limits and signal timing control
to reduce emissions and energy consumption. For example, Zegeye (2011) opti-
mizes T T S, the fuel consumption, and N Ox emissions via MPC. A very similar
1085 approach is presented in Liu et al. (2017). The authors also use MPC to control
both ramp metering and speed limits on a highway section, and the objective
function is the weighted sum of T T S and the total emissions. A specific feature
of this work is that multiple classes of vehicles are considered.
A problem to study in an urban environment is bandwidth maximization
1090 along an artery. Assuming that all the traffic lights have a common cycle, the
problem of bandwidth maximization consist in maximizing the vehicle through-
put along the artery under study, by traffic lights offset control. Usually, the
actuators are only the traffic lights offset, like presented in Mehr et al. (2018) in
which the authors express a nonlinear optimization problem and convert it to a
1095 MILP. The bandwidth maximization problem optimizes the flow of vehicles but
does not explicitly reduce the emissions and energy consumption. Therefore,
De Nunzio et al. (2015) propose a formulation as an optimization problem in
which the objective function contains also terms approximating T T S and energy
consumption. In this work, the authors use coordinated actuators, namely speed
1100 limits and signal timing controls. None of the bandwidth maximization strate-
gies presented is based on a traffic model. Hence, they work best in steady-state
under-saturated traffic conditions.
5.4. Dynamic routing

Another solution to reduce emissions and energy consumption is to use dy-
1105 namic routing. This method consists in redistributing the traffic demand over
35
the network in a more efficient way by controlling the split ratios. In practice,
the controller predicts the optimal routes for the main traffic flow directions,
and the associated recommendations are communicated to the road users by
the mean of in-vehicle devices, radio, or variable message signs (Treiber and
1110 Kesting, 2013a).
In the literature, the control objective of dynamic routing problems is usu-
ally to reach system-optimum or user-equilibrium. The system-optimum corre-
sponds to the minimum T T S and the user-equilibrium is characterized by a den-
sity distribution for which all used routes between the same origin-destination
1115 pair have the same travel time (Xu et al., 2011).
Dynamic routing could also be used to directly reduce emissions and en-
ergy consumption. For example, Luo et al. (2016) propose a real-time en-route
diversion control strategy that minimizes T T S, the total emissions and fuel
consumption. The route recommendation provided by variable message signs
1120 is considered as the control variable. The split ratios are calculated from the
route recommendation considering a drivers’ compliance rate which is supposed
to be known. The route diversion control uses MPC based on a parallel Tabu
Search algorithm.
Emission pricing can also be used as a dynamic routing method aiming at
1125 influencing route selection in order to reduce emissions and energy consumption.
This method can be static or dynamic. Dynamic road pricing studies based on
emissions and energy consumption are reviewed in Wang et al. (2018).
5.5. Mobile actuators

Most of the strategies presented above are motionless in the sense that traffic
1130 lights, ramp metering, message signs and speed limit signs exert commands at a
fixed location. A new approach is to consider mobile actuators, namely vehicles
that could be controlled to have an impact on the surrounding traffic. Typically,
this corresponds to the injection of some autonomous vehicles in the traffic flow
with the objective of stabilizing it.
1135 Stern et al. (2019) present how this method can reduce the emissions of the
whole traffic by dampening stop-and-go waves. To validate this approach, the
authors present the results of field experiments in which vehicle velocity and
acceleration data are collected. These experiments use a single autonomous
vehicle to dampen traffic waves on a ring road with 20 other human-piloted
1140 vehicles. The results are coherent with the simulations of Wu et al. (2018). Yang
and Jin (2014) present a similar control based on inter-vehicle communication.
Autonomous vehicles present opportunities in terms of traffic stabilization,
emissions, and energy consumption. They also induce a smoother driving and
fewer braking events. But the results presented in Stern et al. (2019) hold for
1145 situations with traffic waves only.
A country-level evaluation of the impact of autonomous vehicles on the en-
vironment can be found in Liu et al. (2019). The authors consider different
scenarios regarding the autonomous vehicle penetration rate by 2050.
36
6. Conclusion and outlook
1150 The current situation regarding pollutant emissions and energy consumption
of road transportation is alarming both for environmental and health reasons.
Ecological traffic management appears to be a promising lever in the long-term
to reduce the environmental impact of transportation.
This paper surveys the existing emission and energy consumption models,
1155 as well as the traffic control strategies to reduce them, either by considering
vehicles independently, or by considering traffic flows. The main advantages
and drawbacks of the different approaches are highlighted.
The first step to estimate emissions and energy consumption is to measure,
or simulate, the kinematics of vehicles, that can be either static or dynamic.
1160 Traffic models can be implemented on a microscopic or macroscopic scale. The
complexity of large scale road networks is essentially due to to the processing of
junctions, and the choice of temporal and spatial discretizations, which represent
a crucial point.
From the traffic kinematics, emissions and energy consumption can be esti-
1165 mated using either data or physics-based approaches. Thus, many associations
of models are possible. For complexity reasons, some are more suitable than
others. In order to go large scale, the objective is to find a balance between
accuracy and computation time, which depends mainly on the use of the frame-
work (e.g. compatibility with control methods). For example, a question is
1170 whether the additional complexity introduced by a second-order traffic model
significantly improves the accuracy of a first order model in depicting the traffic
behaviors that impact energy efficiency. Similarly, a microscopic approach to
describe large-scale emissions and energy consumption would provide the best
estimations but it would involve a lot of data that can be difficult to obtain
1175 and process, the need to precisely calibrate the model, and a sharp increase in
computation times. However, this approach can be useful for offline validation
purposes.
Traffic management can be carried out by controlling a single vehicle to
reduce its emissions and energy consumption, or by acting on a large spatial
1180 scale with actuators such as speed limits, traffic lights, dynamic routing or
autonomous vehicles. Usually, a multi-objective optimization is considered to
control the traffic with ecological concerns in order to ensure realistic solutions.
Many articles propose traffic control designs that do not aim at reducing
emissions and energy consumption but more classic metrics such as the distance
1185 traveled, the delays or the total time spent in the networks. These methods can
be adapted to multi-objective control problems considering ecological issues,
which offers promising opportunities in this research field. Also, note that some
control designs aiming at improving traffic fluidity can have a positive impact
on environmental metrics as they reduce the number of stops and accelerations.
1190 Some clear trends can be identified in the ecological approach of traffic con-
trol. For example, autonomous vehicles are considered the next major techno-
logical advance in the transport sector. Not only do they have an important
role to play in road safety, but they can also reduce the impact of transport
37
on the environment by reducing vehicle ownership and improving energy con-
1195 sumption rate (Liu et al., 2019). However, at system-wide level, the effect of
autonomous vehicles on travel demand and energy efficiency is very uncertain
and might increase the total fuel consumption (Brown et al., 2014; U.S. De-
partment of Energy, 2018). Autonomous vehicles are a lever able to influence
and regulate the surrounding traffic. Analyses of the best penetration rate of
1200 connected and automated vehicles in freeway traffic to improve global energy
performance are recently appearing (Rios-Torres and Malikopoulos, 2018), and
further exploration of the effects in an urban environment should be conducted.
Connected vehicles able to communicate with the infrastructure are also
expected to become more numerous, which would considerably increase the
1205 available data. Moreover, computing capabilities have recently been greatly im-
proved. These aspects are expected to improve the efficiency of control strategies
and make microscopic approaches a more interesting option.
Machine learning methods are also a major trend both for estimation and
reduction purposes of emissions and energy consumption. Neural networks are
1210 becoming more and more precise to estimate emissions and energy consumption
from vehicle operating variables. They are a good alternative to physics-based
approaches because of the high non-linearity of emissions and energy consump-
tion. To control the traffic, by any actuator, RL approaches are an important
trend.
1215 MPC approaches seem to be very popular among traffic control methods
as it is compatible with the uncertainties of the traffic models, and it can also
handle non-linear and non-convex optimization.
Some gaps can be identified concerning the ecological approach of traffic
control. One of them is that obtaining large-scale data is difficult because most
1220 vehicles are still not fully connected. The issue of missing data imputation has
been addressed by some authors (Qu et al., 2009).
Moreover, experiments in traffic control are very long and expensive to put
in place. Hence, most approaches are not validated by real experimentation.
However, more and more cities are taking action to reduce pollutant emissions.
1225 For example, many cities are generalizing the speed limit to 30 km/h in most
streets (Bordarie, 2017). Similarly, old diesel vehicles are being banned from
many large cities, especially in Germany (Möhner, 2018). These strong mea-
sures could be associated with a dynamic control of traffic aiming at explicitly
reducing emissions and energy consumption.
1230 Another identified gap is that models intended for large-scale control pur-
poses are limited by computation time. A major issue is their level of detail
(e.g. approximation of the acceleration of macroscopic traffic models, process-
ing of junctions in road networks). This determines the compromise between
accuracy and computation time, which inevitably leads to approximations. In a
1235 hypothetical future in which many vehicles would be connected or autonomous,
the question of data processing from a computational point of view for control
purposes would also certainly arise.
Concerning control strategies, a moot point is to find the metric to optimize
along with the ecological issues. Depending on the objective, many approaches
38
1240 are possible (e.g. minimize the travel duration or distance, homogenize the
density or the speed).
An interesting aspect to study would be the impact of traffic congestion
on emissions and energy consumption. For example, it could be interesting to
analyze in detail the most impacting traffic phenomena on emissions and energy
1245 consumption.
The best models for emission and energy consumption reduction for large-
scale road networks are probably yet to be found. To control the traffic with
ecological concerns, one may explore the use of new actuators, or coordinate
them at a large spatial scale.
1250 Acknowledgements
This project has received funding from the European Research Council
(ERC) under the European Union’s Horizon 2020 research and innovation pro-
gramme (grant agreement 694209).
39
References
1255 Aboudolas, K., Papageorgiou, M., and Kosmatopoulos, E. (2009). Store-and-

forward based methods for the signal control problem in large-scale congested
urban road networks. Transportation Research Part C: Emerging Technolo-
gies, 17(2):163 – 174.
Ahn, K. (1998). Microscopic fuel consumption and emission modeling. Master’s
1260 thesis, Virginia Tech.
Ahn, K. and Rakha, H. (2008). The effects of route choice decisions on ve-
hicle energy consumption and emissions. Transportation Research Part D:
Transport and Environment, 13(3):151 – 167.
Ahn, K., Rakha, H., and Trani, A. (2004). Microframework for modeling of
1265 high-emitting vehicles. Transportation Research Record, 1880:39–49.
Alsabaan, M., Naik, K., Khalifa, T., and Nayak, A. (2012). Applying vehic-
ular networks for reduced vehicle fuel consumption and co2 emissions. In
Intelligent Transportation Systems. IntechOpen.
An, F., Barth, M., Norbeck, J., and Ross, M. (1997). Development of com-
1270 prehensive modal emissions model: operating under hot-stabilized condi-
tions. Transportation Research Record: Journal of the Transportation Re-
search Board, (1587):52–62.
Andersen, O., Jensen, C. S., Torp, K., and Yang, B. (2013). Ecotour: Re-
ducing the environmental footprint of vehicles using eco-routes. In Mobile
1275 Data Management (MDM), 2013 IEEE 14th International Conference on,
volume 1, pages 338–340. IEEE.
Andre, M. and Pronello, C. (1997). Relative influence of acceleration and speed

on emissions under actual driving conditions. International Journal of Vehicle
Design, 18(3-4):340–353.
1280 Aw, A. and Rascle, M. (2000). Resurrection of” second order” models of traffic
flow. SIAM journal on applied mathematics, 60(3):916–938.
Barth, M., An, F., Norbeck, J., and Ross, M. (1996). Modal emissions mod-
eling: A physical approach. Transportation Research Record: Journal of the
Transportation Research Board, (1520):81–88.
1285 Barth, M. and Boriboonsomsin, K. (2008). Real-world carbon dioxide impacts
of traffic congestion. Transportation Research Record: Journal of the Trans-
portation Research Board, (2058):163–171.
Bertsekas, D. P. (1995). Dynamic programming and optimal control, volume 1.
Athena scientific Belmont, MA.
40
1290 Boehme, T., Held, F., Rollinger, C., Rabba, H., Schultalbers, M., and Lampe,
B. (2013). Application of an optimal control problem to a trip-based energy
management for electric vehicles. SAE International Journal of Alternative
Powertrains, 2:115–126.
Bordarie, J. (2017). Public policy of urban mobility: Impact of the history
1295 and practices on young drivers’ social representation of 30 km/hr. Journal of
Nonprofit & Public Sector Marketing, 29(2):211–234.
Boriboonsomsin, K., Barth, M. J., Zhu, W., and Vu, A. (2012). Eco-routing nav-
igation system based on multisource historical and real-time traffic informa-
tion. IEEE Transactions on Intelligent Transportation Systems, 13(4):1694–
1300 1704.
Brown, A., Gonder, J., and Repac, B. (2014). An analysis of possible energy
impacts of automated vehicles. In Meyer, G. and Beiker, S., editors, Road
vehicle automation, Lecture notes in mobility, pages 137–153. Springer, Cham.
Burghout, W. (2004). Hybrid microscopic-mesoscopic traffic simulation. PhD
1305 thesis, KTH.
Canudas-de-Wit, C. (2011). Best-effort highway traffic congestion control via
variable speed limits. In Decision and Control and European Control Confer-
ence (CDC-ECC), 2011 50th IEEE Conference on, pages 5959–5964. IEEE.
Canudas-de-Wit, C. and Ferrara, A. (2018). A variable-length cell transmission

1310 model for road traffic systems. Transportation Research Part C: Emerging
Chapman, L. (2007). Transport and climate change: a review. Journal of
Transport Geography, 15(5):354–367.
Colombo, R. M. (2002). Hyperbolic phase transitions in traffic flow. SIAM

1315 Journal on Applied Mathematics, 63(2):708–721.
Csikós, A., Luspay, T., and Varga, I. (2011). Modeling and optimal control
of travel times and traffic emission on freeways. IFAC Proceedings Volumes,
44(1):13058 – 13063. 18th IFAC World Congress.
Daganzo, C. F. (1994). The cell transmission model: A dynamic representation

1320 of highway traffic consistent with the hydrodynamic theory. Transportation
Research Part B: Methodological, 28(4):269–287.
Daganzo, C. F. (1995). Requiem for second-order fluid approximations of traffic
flow. Transportation Research Part B: Methodological, 29(4):277–286.
De Nunzio, G., Canudas-de-Wit, C., and Moulin, P. (2014). Urban traffic eco-
1325 driving: A macroscopic steady-state analysis. In Control Conference (ECC),
2014 European, pages 2581–2587. IEEE.
41
De Nunzio, G., Canudas-de-Wit, C., Moulin, P., and Di Domenico, D. (2016).
Eco-driving in urban traffic networks using traffic signals information. Inter-
national Journal of Robust and Nonlinear Control, 26(6):1307–1324.
1330 De Nunzio, G., Gomes, G., Canudas-de-Wit, C., Horowitz, R., and Moulin,
P. (2015). Arterial bandwidth maximization via signal offsets and variable
speed limits control. In Decision and Control (CDC), 2015 IEEE 54th Annual
Conference on, pages 5142–5148. IEEE.
De Nunzio, G. and Gutman, P.-O. (2017). An application of shock wave theory
1335 to urban traffic control via dynamic speed advisory. In hEART 2017: 6th
Symposium of the European Association for Research in Transportation.
De Nunzio, G., Thibault, L., and Sciarretta, A. (2017). Bi-objective eco-routing

in large urban road networks. In 2017 IEEE 20th International Conference
on Intelligent Transportation Systems (ITSC), pages 1–7.
1340 Dib, W., Chasse, A., Di Domenico, D., Moulin, P., and Sciarretta, A. (2012).
Evaluation of the energy efficiency of a fleet of electric vehicle for eco-driving
application. Oil & Gas Science and Technology–Revue d’IFP Energies nou-
velles, 67(4):589–599.
Dimitrakopoulos, G. and Demestichas, P. (2010). Intelligent transportation
1345 systems. IEEE Vehicular Technology Magazine, 5(1):77–84.
Dinopoulou, V., Diakaki, C., and Papageorgiou, M. (2006). Applications of the

urban traffic control strategy tuc. European Journal of Operational Research,
175(3):1652 – 1665.
Ericsson, E., Larsson, H., and Brundell-Freij, K. (2006). Optimizing route choice
1350 for lowest fuel consumption – potential effects of a new driver support tool.
Transportation Research Part C: Emerging Technologies, 14(6):369 – 383.
Ferrara, A., Sacone, S., and Siri, S. (2018a). First-order macroscopic traffic
Ferrara, A., Sacone, S., and Siri, S. (2018b). Microscopic and mesoscopic traffic
1355 models. In Freeway traffic modelling and control, chapter 5. Springer.
Ferrara, A., Sacone, S., and Siri, S. (2018c). Second-order macroscopic traffic
Fiori, C., Arcidiacono, V., Fontaras, G., Komnos, D., Makridis, M., Marzano,
V., Mattas, K., Thiel, C., and Ciuffo, B. (2018). Less congestion implies less
1360 energy consumption: Is it really true? Technical report.
Frey, H. C. and Zheng, J. (2002). Probabilistic analysis of driving cycle-
based highway vehicle emission factors. Environmental science & technology,
36(23):5184–5191.
42
Garavello, M., Han, K., and Piccoli, B. (2016). Models for Vehicular Traffic on
1365 Networks.
Gärtner, U., Hohenberg, G., Daudel, H., and Oelschlegel, H. (2004). Devel-
opment and application of a semi-empirical nox model to various hd diesel
engines. In Thermo-and Fluid Dynamic Processes in Diesel Engines 2, pages
285–312. Springer.
1370 Gazis, D. C. and Potts, R. B. (1963). The oversaturated intersection. Technical
report.
Grandinetti, P., Canudas-de Wit, C., and Garin, F. (2018). Distributed optimal
traffic lights design for large-scale urban networks. IEEE Transactions on
Control Systems Technology, (99):1–14.
1375 Greenberg, H. (1959). An analysis of traffic flow. Operations Research, 7(1):79–
85.
Greenshields, B., Channing, W., Miller, H., et al. (1935). A study of traf-
fic capacity. In Highway research board proceedings, volume 1935. National
Research Council (USA), Highway Research Board.
1380 Guanetti, J., Kim, Y., and Borrelli, F. (2018). Control of connected and au-
tomated vehicles: State of the art and future challenges. Annual Reviews in
Control.
Hall, R. (2012). Handbook of transportation science, volume 23. Springer Science

& Business Media.
1385 Han, J., Sciarretta, A., Ojeda, L. L., De Nunzio, G., and Thibault, L. (2018).
Safe-and eco-driving control for connected and automated electric vehicles
using analytical state-constrained optimal solution. IEEE Transactions on
Intelligent Vehicles, 3(2):163–172.
Han, K., Liu, H., Gayah, V. V., Friesz, T. L., and Yao, T. (2016). A robust
1390 optimization approach for dynamic traffic signal control with emission consid-
erations. Transportation Research Part C: Emerging Technologies, 70:3–26.
Hausberger, S. (2009). Emission factors from the model phem for the hbefa
version 3.
Hegyi, A., Hoogendoorn, S. P., Schreuder, M., Stoelhorst, H., and Viti, F.
1395 (2008). Specialist: A dynamic speed limit control algorithm based on shock
wave theory. In 2008 11th International IEEE Conference on Intelligent
Transportation Systems, pages 827–832.
Hegyi, A., Schutter, B. D., and Hellendoorn, H. (2005). Model predictive con-
trol for optimal coordination of ramp metering and variable speed limits.
1400 Transportation Research Part C: Emerging Technologies, 13(3):185 – 209.
43
Helbing, D. (1995). Improved fluid-dynamic model for vehicular traffic. Physical
Review E, 51(4):3164.
Hellström, E., Ivarsson, M., Åslund, J., and Nielsen, L. (2009). Look-ahead
control for heavy trucks to minimize trip time and fuel consumption. Control
1405 Engineering Practice, 17(2):245–254.
Herty, M. and Rascle, M. (2006). Coupling conditions for a class of second-order
models for traffic flow. SIAM Journal on mathematical analysis, 38(2):595–
616.
HomChaudhuri, B., Vahidi, A., and Pisu, P. (2017). Fast model predictive
1410 control-based fuel efficient control strategy for a group of connected vehicles
in urban road conditions. IEEE Transactions on Control Systems Technology,
25(2):760–767.
Hoogendoorn, S. P. and Bovy, P. H. (2001). State-of-the-art of vehicular traffic
flow modelling. Proceedings of the Institution of Mechanical Engineers, Part
1415 I: Journal of Systems and Control Engineering, 215(4):283–303.
Hunt, P., Robertson, D., Bretherton, R., and Winton, R. (1981). Scoot-a traffic
responsive method of coordinating signals. Technical report.
International Energy Agency (2018). World energy outlook 2018.
Jafarmadar, S. (2015). A comparative analysis of two neural network predictions
1420 for performance and emissions in a biodiesel fuelled diesel engine.
(2017). A mesoscopic integrated urban traffic flow-emission model. Trans-
portation Research Part C: Emerging Technologies, 75:45–83.
1425 (2018). Sustainable model-predictive control in urban traffic networks: Effi-
cient solution based on general smoothening methods. IEEE Transactions on
Control Systems Technology, 26(3):813–827.
Jie, L., Zuylen, H. V., Chen, Y., Viti, F., and Wilmink, I. (2013). Calibration
of a microscopic simulation model for emission calculation. Transportation
1430 Research Part C: Emerging Technologies, 31:172 – 184.
Jurik, T., Cela, A., Hamouche, R., Natowicz, R., Reama, A., Niculescu, S.,
and Julien, J. (2014). Energy optimal real-time navigation system. IEEE
Intelligent Transportation Systems Magazine, 6(3):66–79.
Karbowski, D., Kim, N., and Rousseau, A. (2014). Route-based online energy
1435 management of a phev and sensitivity to trip prediction. In Vehicle Power
and Propulsion Conference (VPPC), 2014 IEEE, pages 1–6. IEEE.
Kerner, B. S. and Konhäuser, P. (1993). Cluster effect in initially homogeneous
traffic flow. Physical Review E, 48(4):R2335.
44
Khamis, M. A. and Gomaa, W. (2012). Enhanced multiagent multi-objective
1440 reinforcement learning for urban traffic light control. In 2012 11th Interna-
tional Conference on Machine Learning and Applications, volume 1, pages
586–591. IEEE.
Khondaker, B. and Kattan, L. (2015a). Variable speed limit: A microscopic
analysis in a connected vehicle environment. Transportation Research Part
Khondaker, B. and Kattan, L. (2015b). Variable speed limit: An overview.
Transportation Letters, 7.
Kluge, S., Santa, C., Dangl, S., Wild, S., Brokate, M., Reif, K., and Busch,
F. (2013). On the computation of the energy-optimal route dependent on
1450 the traffic load in ingolstadt. Transportation Research Part C: Emerging
Kubička, M., Klusáček, J., Sciarretta, A., Cela, A., Mounier, H., Thibault, L.,
and Niculescu, S.-I. (2016). Performance of current eco-routing methods. In
Intelligent Vehicles Symposium (IV), 2016 IEEE, pages 472–477. IEEE.
1455 Lejri, D., Can, A., Schiper, N., and Leclercq, L. (2018). Accounting for traffic
speed dynamics when calculating copert and phem pollutant emissions at the
urban scale. Transportation Research Part D: Transport and Environment,
63:588 – 603.
Lelouvier, A., Guanetti, J., and Borrelli, F. (2017). Eco-platooning of au-

1460 tonomous electrical vehicles using distributed model predictive control. pa-
rameters, 2:4.
Lighthill, M. J. and Whitham, G. B. (1955). On kinematic waves ii. a theory of
traffic flow on long crowded roads. Proc. R. Soc. Lond. A, 229(1178):317–345.
Lin, S., De Schutter, B., Xi, Y., and Hellendoorn, H. (2012). Efficient network-
1465 wide model-based predictive control for urban traffic networks. Transportation
Research Part C: Emerging Technologies, 24:122–140.
Lin, S., De Schutter, B., Xi, Y., and Hellendoorn, H. (2013). Integrated ur-
ban traffic control for the reduction of travel delays and emissions. IEEE
Transactions on Intelligent Transportation Systems, 14(4):1609–1619.
1470 Lin, S. and Xi, Y. (2008). An efficient model for urban traffic network control.
IFAC Proceedings Volumes, 41(2):14066–14071.
Liu, F., Zhao, F., Liu, Z., and Hao, H. (2019). Can autonomous vehicle reduce
greenhouse gas emissions? a country-level evaluation. Energy Policy, 132:462
– 473.
1475 Liu, R. and Tate, J. (2004). Network effects of intelligent speed adaptation
systems. Transportation, 31(3):297–325.
45
Liu, S., Hellendoorn, H., and De Schutter, B. (2017). Model predictive control
for freeway networks based on multi-class traffic flow and emission models.
IEEE Transactions on Intelligent Transportation Systems, 18(2):306–320.
1480 Lowrie, P. et al. (1982). The sydney co-ordinated adaptive traffic system (scats)
– principles, methodology, algorithms. In Proceedings of the International
Conference on Road Traffic Signaling, London, UK.
Luo, L., Ge, Y.-E., Zhang, F., and Ban, X. J. (2016). Real-time route diver-
sion control in a model predictive control framework with multiple objectives:
1485 Traffic efficiency, emission reduction and fuel economy. Transportation Re-
search Part D: Transport and Environment, 48:332–356.
Mahler, G. and Vahidi, A. (2012). Reducing idling at red lights based on prob-
abilistic prediction of traffic signal timings. In American Control Conference
(ACC), 2012, pages 6557–6562. IEEE.
1490 Mandl, N. and Pinterits, M. (2018). Annual european union greenhouse gas
inventory 1990–2016 and inventory report 2018.
Mehr, N., Sanselme, M., Orr, N., Horowitz, R., and Gomes, G. (2018). Offset
selection for bandwidth maximization on multiple routes. In 2018 Annual
American Control Conference (ACC), pages 6366–6371. IEEE.
1495 Messmer, A. and Papageorgiou, M. (1990). Metanet: a macroscopic simulation
program for motorway networks. Traffic Engineering & Control, 31:466–470.
Mirchandani, P. and Head, L. (2001). A real-time traffic signal control sys-
tem: architecture, algorithms, and analysis. Transportation Research Part C:
Emerging Technologies, 9(6):415 – 432.
1500 Miyatake, M., Kuriyama, M., and Takeda, Y. (2011). Theoretical study on eco-
driving technique for an electric vehicle considering traffic signals. In Power
Electronics and Drive Systems (PEDS), 2011 IEEE Ninth International Con-
ference on, pages 733–738. IEEE.
Möhner, M. (2018). Driving ban for diesel-powered vehicles in major cities: an
1505 appropriate penalty for exceeding the limit value for nitrogen dioxide?
Nannicini, G., Delling, D., Schultes, D., and Liberti, L. (2012). Bidirectional a*
search on time-dependent road networks. Networks, 59(2):240–251.
Negrenti, E. (1999). The ‘corrected average speed’ approach in enea’s tee model:
an innovative solution for the evaluation of the energetic and environmental
1510 impacts of urban transport policies. Science of the total environment, 235(1-
3):411–413.
Newell, G. F. (1993). A simplified theory of kinematic waves in highway traf-
fic, part i: General theory. Transportation Research Part B: Methodological,
27(4):281–287.
46
1515 Ntziachristos, L., Gkatzoflias, D., Kouridis, C., and Samaras, Z. (2009). Copert:
A european road transport emission inventory model. pages 491–504.
Obodeh, O. and Ajuwa, C. (2009). Evaluation of artificial neural network per-
formance in predicting diesel engine nox emissions. European Journal of Sci-
entific Research, 33(4):642–653.
1520 Osorio, C. and Nanduri, K. (2015). Energy-efficient urban traffic management:

a microscopic simulation-based approach. Transportation Science, 49(3):637–
651.
Ozatay, E., Onori, S., Wollaeger, J., Özgüner, Ü., Rizzoni, G., Filev, D., Miche-
lini, J., and Cairano, S. D. (2014a). Cloud-based velocity profile optimization
1525 for everyday driving: A dynamic-programming-based solution. IEEE Trans-
actions on Intelligent Transportation Systems, 15:2491–2505.
Ozatay, E., Ozguner, U., Michelini, J., and Filev, D. (2014b). Analytical solu-
tion to the minimum energy consumption based velocity profile optimization
problem with variable road grade. IFAC Proceedings Volumes, 47(3):7541–
1530 7546.
Panis, L. I., Broekx, S., and Liu, R. (2006). Modelling instantaneous traffic
emission and the influence of traffic speed limits. Science of the total envi-
ronment, 371(1-3):270–285.
Papageorgiou, M., Hadj-Salem, H., Blosseville, J.-M., et al. (1991). Alinea: A

1535 local feedback control law for on-ramp metering. Transportation Research
Record, 1320(1):58–67.
Pasquale, C., Papamichail, I., Roncoli, C., Sacone, S., Siri, S., and Papageor-
giou, M. (2015). Two-class freeway traffic regulation to reduce congestion
and emissions via nonlinear optimal control. Transportation Research Part
1540 C: Emerging Technologies, 55:85–99.
Payne, H. (1971). Models of freeway traffic and control. In Mathematical Models
of Public Systems. Simul. Counc. Proc.
Pelkmans, L. and Debal, P. (2006). Comparison of on-road emissions with emis-
sions measured on chassis dynamometer test cycles. Transportation Research
1545 Part D: Transport and Environment, 11(4):233 – 241.
Piccoli, B. and Tosin, A. (2009). A review of continuum mathematical models
of vehicular traffic. Encyclopedia of Complexity and Systems Science, pages
9727–9749.
Pisarski, D. and Canudas-de-Wit, C. (2016). Nash game-based distributed con-

1550 trol design for balancing traffic density over freeway networks. IEEE Trans.
Control of Network Systems, 3(2):149–161.
47
Post, K., Kent, J., Tomlin, J., and Carruthers, N. (1984). Fuel consumption and
emission modelling by power demand and a comparison with other models.
Transportation Research Part A: General, 18(3):191–213.
1555 Qi, Y. G., Teng, H. H., and Yu, L. (2004). Microscale emission models incorpo-
rating acceleration and deceleration. Journal of Transportation Engineering,
130(3):348–359.
Qu, L., Li, L., Zhang, Y., and Hu, J. (2009). Ppca-based missing data imputa-
tion for traffic flow volume: A systematical approach. IEEE Transactions on
1560 intelligent transportation systems, 10(3):512–522.
Rai, N. K., Vyas, A., and Singh, S. K. (2017). Street level modeling of pollu-
tants for residential areas. International Journal of Engineering Research and
Application, 7(10):52–57.
Rakha, H., Ahn, K., and Trani, A. (2003). Comparison of mobile5a, mobile6,
1565 vt-micro, and cmem models for estimating hot-stabilized light-duty gasoline
vehicle emissions. Canadian Journal of Civil Engineering - CAN J CIVIL
ENG, 30:1010–1021.
Richards, P. I. (1956). Shock waves on the highway. Operations research,
4(1):42–51.
1570 Rios-Torres, J. and Malikopoulos, A. A. (2018). Impact of partial penetrations
of connected and automated vehicles on fuel consumption and traffic flow.
IEEE Transactions on Intelligent Vehicles, 3(4):453–462.
Schiper, N. (2017). Traffic data sampling for air pollution estimation at different
urban scales. PhD thesis, Université de Lyon.
1575 Sciarretta, A., Nunzio, G. D., and Ojeda, L. L. (2015). Optimal ecodriving con-
trol: Energy-efficient driving of road vehicles as an optimal control problem.
IEEE Control Systems Magazine, 35(5):71–90.
Scora, G. and Barth, M. (2006). Comprehensive modal emissions model (cmem),
version 3.01. User guide. Centre for Environmental Research and Technology.
1580 University of California, Riverside, 1070.
Seredynski, M., Mazurczyk, W., and Khadraoui, D. (2013). Multi-segment green

light optimal speed advisory. In Parallel and Distributed Processing Sympo-
sium Workshops & PhD Forum (IPDPSW), 2013 IEEE 27th International,
pages 459–465. IEEE.
1585 Sobrino, N., Monzón, A., and Hernandez, S. (2014). Reduced carbon and en-
ergy footprint in highway operations: The highway energy assessment (hera)
methodology. Netw Spat Econ, 16:1–20.
48
Stern, R. E., Chen, Y., Churchill, M., Wu, F., Monache, M. L. D., Piccoli, B.,
Seibold, B., Sprinkle, J., and Work, D. B. (2019). Quantifying air quality ben-
1590 efits resulting from few autonomous vehicles stabilizing traffic. Transportation
Research Part D: Transport and Environment, 67:351 – 365.
Stevanovic, A., Stevanovic, J., Zhang, K., and Batterman, S. (2009). Optimizing
traffic control to reduce fuel consumption and vehicular emissions: Integrated
approach with vissim, cmem, and visgaost. Transportation Research Record,
1595 2128(1):105–113.
Sturm, P., Kirchweger, G., Hausberger, S., and Almbauer, R. (1998). Instanta-
neous emission data and their use in estimating road traffic emissions. Inter-
national Journal of Vehicle Design, 20.
Sun, C., Guanetti, J., Borrelli, F., and Moura, S. (2018). Robust eco-driving
1600 control of autonomous vehicles connected to traffic lights. arXiv preprint
arXiv:1802.05815.
Sun, J. and Liu, H. X. (2015). Stochastic eco-routing in a signalized traffic

network. Transportation Research Part C: Emerging Technologies, 59:32 –
47. Special Issue on International Symposium on Transportation and Traffic
1605 Theory.
Tajali, M. and Hajbabaie, A. (2018). Dynamic speed harmonization in urban
street networks. Computer-Aided Civil and Infrastructure Engineering.
Tampère, C. M., Corthout, R., Cattrysse, D., and Immers, L. H. (2011). A

generic class of first order node models for dynamic macroscopic simulation
1610 of traffic flows. Transportation Research Part B: Methodological, 45(1):289 –
309.
Taylor, M. (2000). Network modeling of the traffic, environmental and energy

effects of lower urban speed limits. Road and Transport Research, 9:48–57.
Thibault, L., Degeilh, P., Lepreux, O., Voise, L., Alix, G., and Corde, G. (2016).
1615 A new gps-based method to estimate real driving emissions. In 2016 IEEE
19th International Conference on Intelligent Transportation Systems (ITSC),
pages 1628–1633.
Transport & Environment (2018). Co2 emissions from cars: the facts.
Treiber, M. and Kesting, A. (2013a). Model-based traffic flow optimization. In
1620 Traffic Flow Dynamics, chapter 21. Springer.
Treiber, M. and Kesting, A. (2013b). Trajectory and floating-car data. In Traffic

Flow Dynamics, chapter 2. Springer.
U.S. Department of Energy (2018). Autonomous vehicles: Uncertainties and
energy implications.
49
1625 U.S. Energy Information Administration (2017a). International energy outlook.
U.S. Energy Information Administration (2017b). Study of the potential energy
consumption impacts of connected and automated vehicles.
Van den Berg, M., Hegyi, A., De Schutter, B., and Hellendoorn, H. (2007).
Integrated traffic control for mixed urban and freeway networks: A model
1630 predictive control approach. European journal of transport and infrastructure
research EJTIR, 7 (3).
van Wageningen-Kessels, F., van Lint, H., Vuik, K., and Hoogendoorn, S.
(2015). Genealogy of traffic flow models. EURO Journal on Transportation
and Logistics, 4(4):445–473.
1635 Varaiya, P. (2013). The max-pressure controller for arbitrary networks of sig-
nalized intersections. In Advances in Dynamic Network Modeling in Complex
Transportation Systems, pages 27–66. Springer.
Vreeswijk, J. D., Mahmod, M., and van Arem, B. (2013). Energy efficient traffic
management and control - the ecomove approach and expected benefits. In
1640 13th International IEEE Conference on Intelligent Transportation Systems,
pages 955–961.
Walraven, E., Spaan, M. T., and Bakker, B. (2016). Traffic flow optimization:
A reinforcement learning approach. Engineering Applications of Artificial
Intelligence, 52:203 – 212.
1645 Wang, Y., Szeto, W. Y., Han, K., and Friesz, T. L. (2018). Dynamic traffic
assignment: A review of the methodological advances for environmentally
sustainable road transportation applications. Transportation Research Part
B: Methodological, 111:370–394.
Whitham, G. B. (1974). Linear and nonlinear waves. John Wiley & Sons.
1650 World Health Organization (2016). Ambient air pollution: a global assessment
of exposure and burden of disease.
Wu, C., Bayen, A. M., and Mehta, A. (2018). Stabilizing traffic with autonomous
vehicles. In 2018 IEEE International Conference on Robotics and Automation
(ICRA), pages 1–7. IEEE.
1655 Wu, G., Boriboonsomsin, K., and Barth, M. J. (2014). Development and evalu-
ation of an intelligent energy-management strategy for plug-in hybrid electric
vehicles. IEEE Trans. Intelligent Transportation Systems, 15(3):1091–1100.
Xie, B., Xu, M., Härri, J., and Chen, Y. (2013). A traffic light extension to
cell transmission model for estimating urban traffic jam. In PIMRC, pages
1660 2566–2570.
50
Xu, T. D., Sun, L. J., Peng, Z. R., and Hao, Y. (2011). Integrated route guidance
and ramp metering consistent with drivers’ en-route diversion behaviour. IET
Intelligent Transport Systems, 5(4):267–276.
Xu, Z., Kang, Y., and Lv, W. (2017). Analysis and prediction of vehicle exhaust
1665 emission using ann. In 2017 36th Chinese Control Conference (CCC), pages
4029–4033. IEEE.
Yang, H. and Jin, W.-L. (2014). A control theoretic formulation of green driving
strategies based on inter-vehicle communications. Transportation Research
Part C: Emerging Technologies, 41:48–60.
1670 Yperman, I. (2007). The link transmission model for dynamic network loading.
Zegeye, S., De Schutter, B., Hellendoorn, J., Breunesse, E., and Hegyi, A.
(2013). Integrated macroscopic traffic flow, emission, and fuel consumption
model for control purposes. Transportation Research Part C: Emerging Tech-
nologies, 31:158–171.
1675 Zegeye, S. K. (2011). Model-based traffic control for sustainable mobility.

Zegeye, S. K., De Schutter, B., Hellendoorn, J., Breunesse, E. A., and Hegyi,
A. (2012). A predictive traffic controller for sustainable mobility using pa-
rameterized control policies. IEEE Transactions on Intelligent Transportation
Systems, 13(3):1420–1429.
1680 Zhang, H. M. (2002). A non-equilibrium traffic model devoid of gas-like behav-

ior. Transportation Research Part B: Methodological, 36(3):275–290.
Zhang, Y. and Cassandras, C. G. (2018). The penetration effect of connected
automated vehicles in urban traffic: An energy impact study. arXiv preprint
arXiv:1803.05577.
1685 Zhu, F. and Ukkusuri, S. V. (2014). Accounting for dynamic speed limit con-
trol in a stochastic traffic environment: A reinforcement learning approach.
Transportation Research Part C: Emerging Technologies, 41:30 – 47.
Zu, Y., Liu, C., Dai, R., Sharma, A., and Dong, J. (2018). Real-time energy-
efficient traffic control via convex optimization. Transportation Research Part
51
Available
Available online
online at
at www.sciencedirect.com
www.sciencedirect.com
ScienceDirect
Procedia
Procedia Computer
Computer Science
Science 00 (2018) 000–000
Procedia Computer Science 13400 (2018)
(2018) 000–000
137–144 www.elsevier.com/locate/procedia
Procedia Computer Science 00 (2018) 000–000
The
The 15th
15th International
International Conference
Conference on
on Mobile
Mobile Systems
Systems and
and Pervasive
Pervasive Computing
Computing
(MobiSPC
(MobiSPC
The 15th International Conference 2018)
2018)
on Mobile Systems and Pervasive Computing
(MobiSPC 2018)
A Brief Review of Holonic Multi-Agent Models for Traffic and
A Brief Review of Holonic Multi-Agent
Transportation SystemsModels for Traffic and
Transportation Systems
Igor HAMAN TCHAPPIa,b,∗
a,b,∗, Stéphane GALLANDaa , Vivient Corneille KAMLAcc ,
Igor HAMAN TCHAPPI , Stéphane GALLAND , cVivient Corneille KAMLA ,
Igor HAMAN TCHAPPIa,b,∗Jean Jean ClaudeGALLAND
Claude
, Stéphane KAMGANG
KAMGANG a c
, Vivient Corneille KAMLAc ,
c Belfort, France
Jean Claude
a LE2I, Univ. Bourgogne KAMGANG
a LE2I, Univ. Bourgogne Franche-Comté, UTBM, F-90010
Franche-Comté, UTBM, F-90010 Belfort, France
b Faculty of Sciences, University of Ngaoundere, Ngaoundere, BP : 454, Cameroon
a LE2I, Univ. Bourgogne Franche-Comté, UTBM, F-90010 Belfort, France
c ENSAI, University of Ngaoundere, Ngaoundere, BP : 454, Cameroon
Abstract
Abstract
Abstract
The
The paper
paper presents
presents aa state
state of
of the
the art
art of
of multilevel
multilevel models
models for
for traffic
traffic and
and transportation
transportation modeling
modeling and
and simulation
simulation using
using holonic
holonic
multi-agent
multi-agent approach.
approach. After
Afterofan introduction
anthe
introduction and presentation
and presentation of holonic
of holonic principles,
principles, concepts
concepts and framework,
and framework, the paper
the paper summarizes
summarizes
The paper presents a state art of multilevel models for traffic and transportation modeling and simulation using holonic
research
research works
works on
on traffic
traffic models
models using
using holonic
holonic multi-agent systems. Then aa discussion is
is given, in particular, aa rationale of
of holonic
multi-agent approach. After an introduction andmulti-agent
presentationsystems. Then
of holonic discussion
principles, given,
concepts andin particular, the
framework, rationale holonic
paper summarizes
systems
systems in
in modeling
modeling and
and simulation
simulation of
of traffic
traffic and
and transportation
transportation and
and open
open issues
issues are
are given.
given.
research works on traffic models using holonic multi-agent systems. Then a discussion is given, in particular, a rationale of holonic
systems in modeling
© 2018 The Authors.and simulation
Published of trafficLtd.
by Elsevier and transportation and open issues are given.
©
© 2018
2018 The
The Authors.
Authors. Published
Published by by Elsevier
Elsevier Ltd.
Ltd.
This is
This is an
an open
open access
access article
article under
under the
the CC
CC BY-NC-ND
BY-NC-ND license
license (http://creativecommons.org/licenses/by-nc-nd/3.0/)
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
This
© is an
2018 open
The
Peer-review access
Authors.
under article
Publishedunder
responsibility the
byof CC
Elsevier
the BY-NC-ND
Ltd.
scientific license
committee (http://creativecommons.org/licenses/by-nc-nd/3.0/).
of the 13th International Conference on Future Networks and
Communications,
This is an
Keywords: open
HolonicFNC-2018
access article
Multiagent and
underthe 15th
the
Systems, CC International
BY-NC-ND
Multilevel Model, Conference
license
Traffic, on Mobile Systems and Pervasive Computing, MobiSPC 2018.
Transportation
Keywords: Holonic Multiagent Systems, Multilevel Model, Traffic, Transportation
Keywords: Holonic Multiagent Systems, Multilevel Model, Traffic, Transportation
1.
1. Introduction
Introduction
1. Introduction
The
The increase
increase ofof traffic
traffic demand
demand andand the
the limited
limited capacity
capacity of of the
the infrastructure
infrastructure cause
cause several
several economical,
economical, financial
financial and
and
environment
environment problems
problems related to traffic and transportation in several countries. In order to better manage traffic and
The increase of trafficrelated
demandto and
traffic
theand transportation
limited capacity ofinthe several countries.
infrastructure In order
cause to economical,
several better manage traffic and
financial and
transportation
transportation problems, there
problems,related has
there has been more
been more and more research
and more research on traffic
on traffic modeling
modeling and simulation field. Agents based
environment problems to traffic and transportation in several countries. In and simulation
order to better field.
manage Agents based
traffic and
technologies
technologies is one
one of
of the powerful technologies for the development of
of distributed complex systems. The success of
transportationisproblems, thethere
powerful technologies
has been more and formoretheresearch
development
on traffic distributed
modeling and complex systems.
simulation field.The success
Agents of
based
Agent
Agent based paradigm
based paradigm continuous
continuous growing because this paradigm allows a natural decomposition of the system into
technologies is one of the powerfulgrowing because
technologies for this paradigm allows
the development a natural decomposition
of distributed complex systems.of the
Thesystem
successinto
of
multiple
multiple agents
agents that
that interact
interact with
with each
each other to
to achieve
otherbecause
achieve a desired
desired global goal. Agents based
based technologies are
are widely
Agent based paradigm continuous growing this aparadigm global
allowsgoal. Agents
a natural technologies
decomposition of the systemwidely
into
used
used for
for traffic
traffic and transportation modeling and simulation. AsAs aa particular type of
of agents based technologies, Holonic
multiple agentsand
thattransportation
interact with modeling
each otherandto simulation.
achieve a desired particular typeAgents
global goal. agents based
based technologies,
technologies areHolonic
widely
Multi-Agent System
Multi-Agent System (HMAS)
(HMAS) approach
approach hashas been
been successfully
successfully applied
applied to
to aa wide
wide range
range of
of applications
applications successfully.
successfully.
used for traffic and transportation modeling and simulation. As a particular type of agents based technologies, Holonic
Holonic organizations
Holonic organizations areare among
among thethe successful organizational
organizational models
models that
that have
have been
been introduced inin Multi-Agent
Multi-Agent
Multi-Agent System (HMAS) approachsuccessful
has been successfully applied to a wide range ofintroduced
applications successfully.
System (MAS)
System (MAS) [12].
[12]. The
The term
term “holon”
“holon” waswas originally
originally coined
coined byby the
the Hungarian
Hungarian philosopher
philosopher Arthur
Arthur Koestler
Koestler [15],
[15],
Holonic organizations are among the successful organizational models that have been introduced in Multi-Agent
System (MAS) [12]. The term “holon” was originally coined by the Hungarian philosopher Arthur Koestler [15],
∗ Corresponding
author. Tel.:
Tel.: +237695206801
+237695206801
E-mail
E-mail address:
address: igortchappi@gmail.com
igortchappi@gmail.com
∗ Corresponding author. Tel.: +237695206801
E-mail address:
1877-0509 igortchappi@gmail.com
1877-0509 © © 2018
© 2018 The
TheAuthors.
The Authors.
Authors. Published by
by Elsevier
Published
Published Ltd.
by Elsevier
Elsevier Ltd. Ltd.
This
This is an open access article under the CC BY-NC-ND license
This is
is an
anopen
openaccess article
access under
article the
under CC
the BY-NC-ND
CC BY-NC-ND (http://creativecommons.org/licenses/by-nc-nd/3.0/).
licenselicense (http://creativecommons.org/licenses/by-nc-nd/3.0/)
1877-0509 © 2018
Peer-review under The Authors. Published
responsibility by Elsevier
of the Ltd.
scientific committee of the 13th International Conference on Future Networks and
This is an open access
Communications, article under
FNC-2018 andthe
theCC BY-NC-ND
15th licenseConference
International (http://creativecommons.org/licenses/by-nc-nd/3.0/).
on Mobile Systems and Pervasive Computing, MobiSPC 2018.
10.1016/j.procs.2018.07.154
138 Igor Haman Tchappi et al. / Procedia Computer Science 134 (2018) 137–144
2 Author name / Procedia Computer Science 00 (2018) 000–000
Level 3 Legend
Interaction
Level 2 Holon
Context
Level 1
Level 0
Fig. 1. A nested holarchy with four holarchical levels
basing it on the Greek word “holos” for “whole” and the suffix “-on” that denotes “part”. According to Koestler
[15], a holon is a fractal structure that is stable, coherent, and consisting of several holons acting as substructures. A
biological example is given by Arthur Koestler. For example a human being consists of organs which in turn consist
of cells that can be further decomposed and so on. None of these biological components can be understood without
its sub-components or without the super-component it is part of. In the context of multi-agent systems [12], a holon
is assimilated to an agent that could be composed by other agents. Holonic modeling is used to model the intrinsic
hierarchical nature of the traffic systems. HMAS allows to model multilevel model of traffic by defining several
abstraction levels of system ranging from microscopic to macroscopic level [24].
Three main approaches are presented in literature to model road traffic [24], for instance: microscopic, intermediate
(mesoscopic and hybrid or multilevel) and macroscopic approach. These approaches differentiate the time and space
scales. Each approach has his own advantages and disadvantages. Macroscopic level requires a high number of entities
and generally is applied on highway. While, microscopic level requires a high computational cost and generally is
applied on small urban area scale with a high degree of accuracy. Hybrid models or multilevel models were developed
in order to study the situations in which macroscopic and microscopic models are not well adapted like modeling
and simulation of large scale traffic [17]. Multilevel approach combines the advantages of macroscopic level and
microscopic level but it may be difficult to realize [10]. The paper presents a state-of-the-art of the multilevel models
of traffic and transportation using holonic multi-agent system (HMAS).
The paper is organized as follows: in Section 2, a brief description of HMAS, principles, concepts and framework
is presented. Section 3 summarizes research works on modeling and simulation of traffic using HMAS. In Section 4,
motivations, drawbacks and open issues of HMAS in traffic and transportation are given. Finally, Section 5 gives a
conclusion and future works.
2. Holonic Multiagent Systems
The organizational approach allows to model complex systems. Its strengths are modularity, multiple architectures,
heterogeneity of languages, application safety and reusability [7]. Organizational approach and holonic modeling
are a suitable approach to reduce the complexity of complex systems by defining several abstraction levels [10].
Holonic modeling is used to model the intrinsic hierarchical nature of systems. Holonic organizations have proven to
be effective solutions to several problems associated with hierarchical and self-organizing structures and have been
successfully applied in a wide range of domains including multi-agent systems [12], philosophy [27], manufacturing
systems [14], transportation [22] to name a few. A holon, according to Koestler [15] is defined as simultaneously a
whole and a part of the whole. Thus, it can be made up of other holons, strictly meeting three conditions: being stable,
having a capacity for autonomy and being able to cooperate. One of the most interesting properties of the holonic
systems, which is the essence of their complexity, is that a holon can be both an entity and an organization. The
holons are therefore stable and self-similar structures. A holarchy is shown on Fig. 1.
Igor Haman Tchappi et al. / Procedia Computer Science 134 (2018) 137–144 139
Author name / Procedia Computer Science 00 (2018) 000–000 3
3. Paper Summaries
The modeling approaches and findings described in the most relevant studies are reviewed in the following. Al-
though holonic approach has been successfully applied to a wide range of applications [12, 27, 14, 22, 10]. This
review concentrates on HMAS simulation models for traffic and transportation. To this end, the paper focus on the
review of HMAS modeling frameworks, holonic models of individuals (pedestrians and vehicles), holonic models of
the environment (dedicated to traffic) and holonic models of traffic signals control.
3.1. Frameworks for Holonic Systems Modeling
Most of the works related to modeling and simulation of traffic with HMAS are based on CRIO metamodel [21, 3].
The CRIO (Capacity - Role - Interaction - Organization) model is one of the framework to design holonic multi-agent
system. CRIO model is based on four main interrelated concepts:
• Capacity: A capacity is a description of a know-how/service. This description contains at least a name identi-
fying the capacity and the set of its input and output variables which may have default values.
• Organization: A set of roles and their interactions pattern define an organization in a specific domain. The
concept of organization combines roles and their interactions.
• Role: A role is the abstraction of a behavior in a certain context and confers a status within the organization.
Roles may interact with other roles defined in the same organization.
• Interaction: An interaction links two roles in a way that an action in the first role produces a reaction in the
second.
The CRIO metamodel aims at providing a full set of abstractions to model MAS and HMAS under an organi-
zational perspective. CRIO uses the development approach defined in the Model Driven Architecture (MDA) [19].
The elements of CRIO are organised in three different domains: (i) The Problem Domain (CIM – Computation Inde-
pendent Model) deals with the user’s problem in terms of requirements, organizations, roles and ontologies; (ii) The
Agency Domain (PIM – Platform Independent Model) addresses the holonic solution to the problem described in the
previous domain; (iii) Finally, the Solution Domain (PSM – Platform Specific Model) describes the structure of the
code solution in the chosen implementation platform.
CRIO has been extended to K-CRIO by Lin et al. [16] in order to model human activities. K-CRIO uses a semantic
layer for collaborative softwares in order to enhance Knowledge Management within the targeted organizations.
CRIO has also been extended to NCRIO by Missaoui et al. [18] in order to include norms within the metamodel.
NCRIO is used to design a normative holonic multi-agent system. This metamodel retains the properties of the holonic
multi-agent systems and adds normative concepts (Norms and Contracts) to maintain social control in these systems.
3.2. Models of Individuals in Traffic using HMAS
This section presents the modeling of the individual behaviours, based on HMAS and the associated outcomes.
Hans-Jürgen et al. [13] propose TELETRUCK in order to support dispatch officers in route planning, fleet manage-
ment, and driver scheduling. TELETRUCK model is among the first HMAS model in transportation. TELETRUCK
model uses holonic multi-agent system in order to achieve a flexible, structured resource management in the planning
process. The main components of TELETRUCK are driver, truck with loading space, truck without loading space,
truck tractor, trailer, semitrailer, chassis, container. One of the main ideas underlying the TELETRUCK approach is
to model his physical components explicitly by basic agents. The agents collaborate closely and join together form-
ing holonic agents that act in a corporated way. The composed agents represent the physical transportation entities
(e.g. road trains or articulated vehicles together with their drivers), which are able to execute the orders based on
an integration with modern telecommunication facilities. The authors propose an algorithm and protocols in order to
deal with communication, coordination, and resource control of their holonic agents. This algorithm starts with an
initial sub-optimal schedule and improves the solution at every negotiation cycle. TELETRUCK increases efficiency,
transparency and leads to a reduction of traffic congestion, air pollution, and energy consumption.
Gaud et al. [11] propose a holonic organizational multilevel model for real-time simulation of pedestrians by
exploiting the hierarchical and distributed properties of the holarchies. Their model is bottom up and it’s a force based
model. The authors estimate the deviation of simulation accuracy between two adjacent levels, based on the use of
physics-based indicators. These indicators allow to dynamically determine the most suitable level for each entity in
application to maintain the best compromise between simulation accuracy and available resources. To evaluate the
accuracy of their model, three measurements inspired by different energy values are proposed : Kinetic energy Eci ,
Goal potential energy E pgi and Constraints potential energy E pci . The authors define the global energy of a holon i
by Ei = Eci + E pgi + E pci . This energy represents the current state of a holon and thus can be used to determine
the deviation of the simulation accuracy between two adjacent levels. In this model, the pedestrian behavior of a
super-holon is the same of its sub-holons but perceptions/actions are aggregated. Affinity between two holons i and
j is defined according to three main functions : the distance between holon objectives, the distance between holon
location, and the energetic affinity such that Af fener (i, j) = 1/Ein − Enj . In experimentation, the authors validate their model
saying that scheduling holarchy implies a reduction of computational cost.
Rodriguez et al. [22] propose a framework for modeling and simulation of an important industrial plant (Peugeot car
manufacturer at Sochaux, France). The plant, produces over a 1, 700 automobiles per day, counts 19, 000 employees,
occupies a surface of 250 hectares and contains even an internal railway. This work presents a model based on a multi-
view analysis of a plant: Traffic Flow View (describing the structure and decomposition of the environment and the
way vehicles interacted with this environment) and View Family of Building (identify the product exchange among
buildings of the plant based on the available traffic information). Each view is modeled in terms of holons reducing
the complexity and size of the model. The integrated model of the simulator is the fusion of the view models. Plant
is recursively decomposed in order to build holons and holarchy. They simulate traffic containing different types of
vehicles (cars, trucks, etc.) and estimate congestion in the plant. Vehicles are composed of three fairly independent
modules: physical characteristics (contains physic related contents like maximal speed, maximal acceleration etc.),
driving logic (encapsulates the actual behavior of the agent, including route planning) and control logic (provides a
facade that maps driving logic commands, like speedup). The authors define satisfaction in order to allow holons to
dynamically change their roles. To validate their model, authors present a 3D representation of the simulated world
on a virtual reality platform.
3.3. Models of Road Network and Traffic Control using HMAS
Abdoos et al. [2] propose a method to construct the holarchy for a multi-agent urban area network based on graph
theory in order to recognize and form the holons. They use graph based modeling approach to group a population
of agents with a greedy method. Agents are modeled using an undirected and weighted graph, in which the weights
denote the degrees of the dependencies between the agents. In order to evaluate the consistency of their method and to
build the holarchy, the authors propose a quality measure based on the dependencies between the agents. It is a bottom
up method. In the case study of the paper, the crossroads are modeled by nodes of a graph, and the roads by edges. The
model has been tested on a road network with 18 crossroads and 39 two-way links, using the Aimsun traffic simulator.
The experimental results show that the average delay time is reduced in comparison with an homogeneous method.
Abdoos et al. [1] extend the previous works by introducing a holonic Q-learning algorithm to control the signals.
In this model, each agent represents a crossroad or traffic signal. The model uses HMAS to model a large number
of signals’ controllers agents. In order to reduce the complexity of traffic control for a large network, the authors
divide the network into sub-systems, each assigned to a holon. In the case study, the method has been applied on
a large network that contains 50 crossroads and 112 two-way links. The experimental results show that the holonic
Q-learning prevents the network to be over-saturated. While, it causes less average delay time and higher flow rate.
Esmaeili et al. [6] propose a method, inspired by social networks, to build the initial holonic structures of a multi-
agent network with a bottom-up approach. Their paper presents a holonification method for a population of intercon-
nected homogeneous agents. The importance of the agents in this approach in the flat network structure is evaluated
using the centrality concept in social network. The prerequisite of their method is an un-weighted undirected multi-
agent network model. A networked multi-agent system is presented as a graph such that vertices are the agents and
edges are the interaction relationship between agents. Having built the initial holons, the other agents of the network
try to join these holons to form a multi-level holarchy. They use urban traffic signal control to evaluate the quality of
the constructed holons. This method has been tested on an urban traffic road network containing 25 crossroads and
Table 1. Comparison between HMAS models in traffic modeling and simulation field
Models Building Method Clustering Method Goal of the model
Gaud et al. [11] Bottom up (holons grouping) Holons affinity Pedestrian simulation; compromise simulation accuracy /
computational resources
TELETRUCK [13] Bottom up (agents unification) Components matching, trans- Dynamic planning and on-line optimization of transport;
portation task requirements Route planning; Fleet management; Driver scheduling
Rodriguez et al. [22] Top down (plant decomposi- Satisfaction based on task Modeling and simulation of PSA Peugeot Plant; chain pro-
tion) progress; holons affinity duction optimization; traffic simulation
Esmaeili et al. [6] Bottom up (joining basic Network position, neighbors Modeling road network; Traffic planning; Intelligent traf-
holons) eligibility fic control
Abdoos et al. [2] Bottom up merging nodes Quality measure based on de- Modeling road network; Traffic planning; Intelligent traf-
pendency of holons fic control
Abdoos et al. [1] Bottom (merging nodes, traffic Quality measure on depen- Modelling a large traffic network; Development of a hi-
network partitioning) dency of holons erarchical control strategy; Traffic planning; Learn signal
timing; Intelligent traffic control
Galland and Gaud [8], Galland Top down (topological decom- Place, zone Environment model for crowds simulation; Agentization
et al. [9] position of environment) of the environment model to support multilevel simulation
Salido et al. [23], Versteegh Top down Shortest path Environment simulation of GATS in a manufacturing
et al. [25] plant; Route planning
31 connecting roads. They use average traffic delay time to compare their method and the one of Abdoos et al. [2],
saying that their method causes relatively lower delay time for the vehicles in the traffic network, and postpones the
saturation time of the traffic.
3.4. HMAS Models of the Environment
The environment is an important part of the system that should be studied in details [26]. Several works dedicated
to the environment model in traffic modeling and simulation have been proposed in literature.
Galland and Gaud [8], Galland et al. [9] propose a holonic model of a physical environment for the simulation
of crowds in virtual 3D buildings. The environment model are agentified and therefore, the term agent refers in their
papers to the agents, which are supporting the environment model. The method is a recursive top down topological
decomposition of environment in places (a place is a semi-closed spatial area bounded by static objects). They propose
three main families of energy indicators for evaluating the decomposition process: the mass of a zone which indicates
the importance of a place of the environment, the structural depth describes the minimum or the maximum depth of
the decomposition of a zone, and the resource constraint, which describes the limits of the available resources for a
place to achieve its simulation. The experimentation’s goal for this model is to ensure that the model provides similar
results than other crowd simulation within airports, and to evaluate the impact of the approximation that is applied by
the use of the energy indicators. To this end, the model is successfully applied to the simulation of two airports halls
around 0.25 km2 , which are separated by gates. Experiments permit to evaluate the impact of the multilevel simulation
on the simulation results, and the gain in terms of computational costs.
Versteegh et al. [25] present a distributed architecture for modeling Global Automated Transportation System
(GATS) and the distributed algorithm for simulating it. They focus for that on holonic multi-agent system. The pro-
posed algorithm ensures privacy for each holon. Salido et al. [23] extend this model by introducing an holonic sim-
ulation environment for the GATS in order to manage the dynamic events within this environment. This approach is
designed for route planning in the context of a automated transport system. There are different transportation vehicles
(guided vehicles, trucks, lift trucks and cars) with different features that make the transportation planning a complex
problem. To overcome this issue, the model solves the shortest path between two points in the GATS, and guides the
different vehicles in the transportation system. They applied the model to a manufacturing plant. The authors notice
that the holon reorganization significantly decreased costs of solutions for cases with narrow time windows.
3.5. Summarized Overview
In order to allow a better comparison and a brief overview of the different modeling approaches, the presented
papers are shortly summarized in Table 1.
4. Discussion
4.1. Motivations for HMAS in Traffic
Traffic system is a complex system, and it is a large scale phenomenon [24]: (i) traffic is composed by heterogeneous
entities; (ii) the number of entities composing the traffic is very high and the interactions between these entities are
non-linear; (iii) traffic is geographically and fundamentally a distributed phenomenon; (iv) there are several levels of
detail for traffic observation. HMAS is a suitable tool to deal with this high complexity of traffic:
• Since traffic is geographically and fundamentally a distributed phenomenon, HMAS allows to model large scale
traffic [10].
• HMAS allows to divide the system in perspectives, or views, in order to reduce the number of involved entities,
and the complexity of the process [22].
• HMAS allows to represents a complex system under several views, ranging from microscopic to macroscopic
approaches. In general, the multi-level modeling of road traffic combine simultaneously different road traffic
models in one model [17, 20]. Each model is associated to a part of the road network. The goal of this hy-
bridization is to manage the transition of the models at the border. Therefore, most of the multilevel models
focus only on two abstraction levels (micro and macro, macro and meso, meso and micro). As stated before,
one particularity of HMAS in multi-level modeling of road traffic is that HMAS can combine the three main
abstraction levels (micro, meso and macro) in the same model according to a point of view.
• Although, it is widely recognized that the presence of groups influences microscopic and aggregated pedestrian
or vehicles dynamics, a precise characterization of the phenomenon still calls for evidences and insights [4].
Several works analyzed the influence of the group behavior on an individual. However, there is still a need for
additional insights, for instance on the spatial patterns assumed by the groups in their movement, and in general
on the interaction among different factors influencing overall vehicles dynamics [4]. A holarchy is an effective
solution to link micro and macro levels in order to understand well the influence of micro on the macro level.
• HMAS allows to change dynamically the modeling level (micro, meso1 ,...,meson , macro) according to several
criteria like visualization or computational resources [24, 1]. HMAS enables maintaining the best compromise
between the simulation accuracy and the available resources [10]. Moreover, with HMAS, abstraction levels are
not fixed priori, but depend on the application, and may be dynamically selected.
• The works presented in Section 3 show that HMAS can reduces simulation’s computational costs, i.e. increases
efficiency, and leads to a reduction of traffic congestion, air pollution, and energy consumption, etc.
4.2. Methodology of HMAS in Traffic
To model traffic using HMAS in literature, two main approaches are presented:
• Gaud et al. [10] and Galland and Gaud [8]: The goal of this approach is to find the same behavioral model
for both the individuals and the groups. Considering that the individuals and the groups have a similar behavior,
the self-similarity of the behaviors is directly exploited at different abstraction levels to build the holarchy, and
to simulate the system at different levels. Although, this approach may ease the simulation process, his main
drawback is related to the difficulties for finding the same behavior for both the individuals and the groups.
Indeed, the individuals and the groups may not use the same time scale and features variables.
• Abdoos et al. [1] and Esmaeili et al. [5]: In this approach, the life-cycle of HMAS consists of two stages:
building the initial holarchy and controlling its structure against internal, and external stimuli during its lifetime.
The initial holarchy represents the structure configuration of HMAS at time t = 0, while control structure against
internal and external stimuli represents the life of HMAS at time t > 0. Although, this approach may ease the
building of HMAS (initial holarchy). Its main drawback is to manage self-organization over time particularly
for a highly distributed and open system like a traffic system.
As shown in Table 1, building the holarchy can be made with the bottom up or top down approaches. In HMAS
top down approach reduces the modeling complexity for a system (a holon with a complex objective can decided to
decompose into sub-holon and so on). Otherwise, bottom up approach can leads to emergent properties of the system.
The holons interacts together to achieve a complex behavior.
4.3. Drawback of HMAS in Traffic
Section 3 shows that most of HMAS models in traffic focus on road networks, environment and traffic control. The
common feature of these systems is that there are little evolutive (static) in terms of extension. There are few works
that focus on modeling of individuals, e.g. pedestrians, or vehicles. The models of the individuals presented in this
paper are application dependent (pedestrians in airport, vehicles in plant). This can be explained by the fact that the
modeling of a large, open and dynamic system with HMAS is still a challenge. In fact, if a system is too dynamic,
HMAS spends more time for building the holarchy than running the system itself.
Another drawback of HMAS, which is one of its strengths, is the homogeneity of the holons. The principle of
building holarchy is based on the homogeneity (affinity, satisfaction, quality measure etc.) of the holons’ behaviors.
In traffic, the behaviors of the drivers are generally not homogeneous.
Finally, visual tools representing the traffic flows using HMAS are not provided.
4.4. Open Issues
Several open issues are presented in this section regarding the modeling and simulation of traffic flow using HMAS.
• Sub-microscopic models: HMAS focus only on micro, meso, macro levels. However, it is interesting to focus
on the submicroscopic level also. In this case, it will be possible to include the behaviors of the entities’s
components, e.g. the mechanical components of a car, in order to simulate the physical motion of the agents in
a more realistic way.
• Integration of norms: In traffic application, the integration of norms may be an added-value in order to build the
holarchy following application-dependent knowledge and rules. This issue may be solved by using the NCRIO
framework [18]. This approach should enable the integration of the traffic regulations within the HMAS.
• Take into account the driver behavior: Several works focus on the driver behavior, but none using HMAS to
the best of our knowledge. It is interesting to consider driver behaviors in HMAS, especially when facing the
self-similarity of the holons’ behaviors.
• Modeling large scale, distributed and open system: Since traffic is open and highly dynamic, HMAS can be
apply in situation when traffic is low dynamic like peak hour, congested traffic, platoon, convoy. This affirmation
needs a to be validated and proven.
5. Conclusion
This paper contains a review of relevant papers in the field of holonic multi-agent systems and traffic modeling
and simulation. Table 1 summarizes the core characteristics of the cited works, and highlights the similarities and
differences between the approaches. In Section 4, some of the current problems facing with HMAS are identified.
HMAS research methodology requires further effort (in modeling of individuals) and a consolidation of the approaches
pursued by the different research groups, especially those focus on the environment and traffic control is needed.
Future works include modeling large scale, distributed and open traffic system based on HMAS. Others multilevel
indicators, which support the spatio-temporal properties of the system and the driver behaviors, will be studied and
proposed.
References
[1] Abdoos, M., Esmaeili, A., Ana, L.C., B., 2013. Holonic multi-agent system for traffic signals control. Engineering Applications of Artificial
Intelligence, Elsevier .
[2] Abdoos, M., Esmaeili, A., Mozayani, N., 2012. Holonification of a network of agents based on graph theory, in: Jezic, G., Kusek, M., Nguyen,
N.T., Howlett, R.J., Jain, L.C. (Eds.), Agent and Multi-Agent Systems. Technologies and Applications, Springer Berlin Heidelberg, Berlin,
Heidelberg. pp. 379–388.
[3] Cossentino, M., Gaud, N., Galland, S., Hilaire, V., Koukam, A., 2007. A holonic metamodel for agent-oriented analysis and design, in: Mařı́k,
V., Vyatkin, V., Colombo, A.W. (Eds.), Holonic and Multi-Agent Systems for Manufacturing, Springer Berlin Heidelberg, Berlin, Heidelberg.
pp. 237–246.
[4] Crociani, L., Gorrini, A., Feliciani, C., Vizzari, G., Nishinari, K., Bandini, S., 2017. Micro and macro pedestrian dynamics in counterflow: the
impact of social groups. CoRR abs/1711.08225. URL: http://arxiv.org/abs/1711.08225, arXiv:1711.08225.
[5] Esmaeili, A., Mozayani, N., Motlagh, M., Matson, E., 2017. A socially-based distributed self-organizing algorithm for holonic multi-agent
systems: Case study in a task environment. Cognitive Systems Research, Elsevier , 21–44.
[6] Esmaeili, A., Mozayani, N., Motlagh, M.R.J., 2014. Multi-level holonification of multi-agent networks, in: 2014 Iranian Conference on
Intelligent Systems (ICIS), IEEE. pp. 1–5. doi:10.1109/IranianCIS.2014.6802553.
[7] Ferber, J., Gutknecht, O., Michel, F., 2008. Agent/Group/Roles: Simulating with Organizations, in: J.P., M. (Ed.), ABS’03: Agent Based
Simulation, Montpellier (France). p. 12. URL: https://hal-lirmm.ccsd.cnrs.fr/lirmm-00269714. april 28-30.
[8] Galland, S., Gaud, N., 2014. Holonic model of a virtual 3d indoor environment for crowd simulation, in: Proceedings of the International
Workshop on Environments for Multiagent Systems (E4MAS14), Springer.
[9] Galland, S., Gaud, N., Demange, J., Koukam, A., 2014. Multilevel model of the 3d virtual environment for crowd simulation in buildings, in:
The 1st International Workshop on Agent-based Modeling and Simulation of Cities (AgentCities14), Procedia Computer Science.
[10] Gaud, N., Galland, S., Gechter, F., Hilaire, V., Koukam, A., 2008. Holonic multilevel simulation of complex systems: Application to real-time
pedestrians simulation in virtual urban environment. Simulation Modelling Practice and Theory, Elsevier , 1659–1676.
[11] Gaud, N., Gechter, F., Galland, S., Koukam, A., 2007. Holonic multiagent multilevel simulation application to real-time pedestrians simulation
in urban environment, in: Proceedings of the 20th International Joint Conference on Artifical Intelligence, Morgan Kaufmann Publishers Inc.,
San Francisco, CA, USA. pp. 1275–1280. URL: http://dl.acm.org/citation.cfm?id=1625275.1625482.
[12] Gerber, C., Siekmann, J., Vierke, G., 1999. Holonic multi-agent systems. Research Center for Artificial Intelligence (DFKI) Tech. Rep. 681,
German .
[13] Hans-Jürgen, B., Klaus, F., Gero, V., 2000. Holonic transport scheduling with teletruck. Applied Artificial Intelligence: An. International
Journal .
[14] Hendrik, Van, B., JoWyns, Paul, V., Luc, B., Patrick, P., 1998. Reference architecture for holonic manufacturing systems: Prosa. Computers in
Industry , 255–274.
[15] Koestler, A., 1967. The Ghost in The Machine. Hutchinson.
[16] Lin, Y., Hilaire, V., Gaud, N., Koukam, A., 2011. K-crio: An ontology for organizations involved in product design, in: Cherifi, H., Zain,
J.M., El-Qawasmeh, E. (Eds.), Digital Information and Communication Technology and Its Applications, Springer Berlin Heidelberg, Berlin,
Heidelberg. pp. 362–376.
[17] Mammar, S., Mammar, S., Lebacque, J., 2006. Highway traffic hybrid macro-micro simulation model, in: 11th IFAC Symposium on Control
in Transportation System.
[18] Missaoui, E., Mazigh, B., Bhiri, S., Hilaire, V., 2017. Ncrio: A normative holonic metamodel for multi-agent systems, in: Martı́nez de Pisón,
F.J., Urraca, R., Quintián, H., Corchado, E. (Eds.), Hybrid Artificial Intelligent Systems, Springer International Publishing, Cham. pp. 638–649.
[19] Object Management Group (OMG), 2003. Mda guide. v1.0.1, OMG/2003-06-01 .
[20] Poschinger, A., Kates, R., Keller, H., 2002. Coupling of concurrent macroscopic and microscopic traffic flow models using hybrid stochastic
and deterministic disaggregation. Transportation and Traffic Theory for the 21st century .
[21] Rodriguez, S., Hilaire, V., Koukam, A., 2005. Formal specification of holonic multi-agent systems framework, in: Sunderam, V.S., van Albada,
G.D., Sloot, P.M.A., Dongarra, J. (Eds.), Computational Science – ICCS 2005, Springer Berlin Heidelberg, Berlin, Heidelberg. pp. 719–726.
[22] Rodriguez, S., Hilaire, V., Koukam, A., 2007. Towards a holonic multiple aspect analysis and modeling approach for complex systems:
Application to the simulation of industrial plants. Simulation Modelling Practice and Theory, Elsevier , 521–543.
[23] Salido, M.A., Peinado, M., Giret, A., 2011. A holonic simulation environment for smart transportation systems. International Journal of Pro-
duction Research 49, 1425–1439. URL: https://doi.org/10.1080/00207543.2010.519489, doi:10.1080/00207543.2010.519489,
arXiv:https://doi.org/10.1080/00207543.2010.519489.
[24] Tchappi, I., Kamla, V., Galland, S., Kamgang, J., 2017. Towards an multilevel agent-based model for traffic simulation, Procedia Computer
Science. pp. 887–892.
[25] Versteegh, F., Salido, M.A., Giret, A., 2010. A holonic architecture for the global road transportation system. Journal of Intelligent Manufac-
turing 21, 133–144. URL: https://doi.org/10.1007/s10845-008-0165-6, doi:10.1007/s10845-008-0165-6.
[26] Weyns, D., Parunak, H.V.D., Michel, F., Holvoet, T., Ferber, J., 2006. Environments for Multiagent Systems State-of-the-Art and Research
Challenges, in: Third International Workshop (E4MAS 2006). Springer, Hakodate, Japan. volume 4389 of Lecture Notes in Artificial Intelli-
gence, pp. 1–47.
[27] Wilber, K., 1995. Sex, ecology, spirituality: The spirit of evolution, Shambhala, Halifax.
LAPORAN TUGAS AKHIR
ANALISIS KINERJA SIMPANG BERSINYAL DAN RUAS JALAN SETELAH
DITERAPKAN SISTEM SATU ARAH PADA JALAN DEWI SARTIKA DI
DEPOK
Diajukan untuk memenuhi syarat mata kuliah Tugas Akhir pada Program Sarjana
Strata 1 (S1)
Disusun Oleh:
Rizky Wahyu Pangestu
NIM : 41116320030
PROGRAM STUDI TEKNIK SIPIL
FAKULTAS TEKNIK
UNIVERSITAS MERCU BUANA
BEKASI
2018
1
http://digilib.mercubuana.ac.id/
ScienceDirect
Procedia Computer Science 96 (2016) 1637 – 1646
20th International Conference on Knowledge Based and Intelligent Information and Engineering
Systems, KES2016, 5-7 September 2016, York, United Kingdom
Traffic Flow Optimization on Freeways

Bruno Carlo Rampinelli Rotaa*, Milan Simica
a
RMIT University, 124 La Trobe St, Melbourne 3000, Australia
Abstract
Intelligent Transportation Systems (ITSs) are advanced technologies and applications integrated in transportation networks with
the aim of creating a safe and efficient transportation environment. ITSs are developing both in the cities, that will become smart
cities, and on freeways, by designing smart roads. Focusing on freeways, tollbooths currently represent serious challenges in
many countries. Most commonly used payment toll systems create traffic congestion with the consequent increase of traffic
pollution, time delays and negative results in terms of vehicular throughput and freeways’ efficiency. ITSs integrated with new
arriving Unmanned Ground Vehicles (UGVs) and novel toll collections designs, such as Open Road Tolling (ORT), could
remove, or at least, mitigate these bottlenecks. This paper presents traffic flow models and algorithms implemented in IBM ILOG
CPLEX Optimization Studio. Modeling and comprehensive investigation, using real data, brings advantages in terms of traffic
throughput and optimization.
©©2016
2016The
TheAuthors.
Authors. Published
Published by Elsevier
by Elsevier B.V.B.V.
This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of KES International.
Peer-review under responsibility of KES International
Keywords: Throughput Optimisation; Traffic flow; Open Road Tolling; VANET
1. Introduction
Traffic congestions on the freeways are problems that arise for different reasons. Intelligent Transportations
Systems (ITSs) can reduce them and mitigate the negative consequences2. ITSs integrated with new technologies can
benefit transportation networks in terms of safety improvements3, traffic flow maximization4 and gas pollution
reduction2. According to number of studies5,6, tollbooths represent one of the major causes of these problems. We
could use different mathematical models7-9 to describe these bottlenecks. Different technologies have been
developed, in recent years, in order to increase freeways’ performances and offer better level of service (LOS)6.
*Corresponding author. Bruno Carlo Rampinelli Rota. Tel.: +61 541 110 604.
E-mail address: s3458554@student.rmit.edu.au;
1877-0509 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of KES International
doi:10.1016/j.procs.2016.08.211
1638 Bruno Carlo Rampinelli Rota and Milan Simic / Procedia Computer Science 96 (2016) 1637 – 1646
Moreover, new transportation modes, such as applications of Unmanned Ground Vehicles (UGVs), in the near
future, integrate their technology with recent tollbooths designs, offering comprehensive solutions10. Integrating
these technologies will bring important advantages in traffic control and traffic management, creating faster and real-
time transportation networks11. ITSs, new technologies, transportation modes and infrastructure are components of
the transportation network12, which allow communication and information exchanges between elements inside its
environment3. Vehicular ad Hoc Network (VANET) is a computer network associated with moving vehicles and
road infra-structure. For example, vehicles can already communicate using Dedicated Short Range Communications
(DSRCs) in order to improve transportation systems13 performances. VANET, as an ad-hoc computer network3 of
fast moving nodes, i.e. network elements, is able to support communication and enables guidance of the vehicles that
are in its range while moving with high speeds, as defined by Institute of Electrical and Electronics Engineers
(IEEE) standards. IEEE 802.20 supports various vehicular mobility classes, up to 250 Km/h in a Metropolitan Area
Network (MAN) environment. This is just an extreme case that shows capabilities of the new technologies,
especially in the communications domain. Down to the low speeds, it is evident that integrating these technologies in
freeways’ bottlenecks, such as tollbooths, will contribute to the creation of safer and better performing traffic on the
roads.
Nomenclature
ȡ traffic density
ȡ0 initial density
ȡmax critical density
C0 signal speed
L number of lanes distinguished by the payment method
N total number of vehicles in the time window
q traffic flow
vff free flow speed (i.e. the maximum allowed speed)
T time window
t time
v speed
vff free flow speed
yi Boolean decision variable. Equals to 1 if the i-th vehicle passes through the payment toll in T; 0 otherwise
x space
Z decision variable: it represents the sum of all i-th vehicles that pass through the payment toll in time T
2. Freeways and toll payments designs
It is important to define different types of toll payment methods adopted in freeways. First of all, toll roads can be
categorized according to their payment systems, as explained here:
x Close Systems are toll roads with entry and exit barriers. Payment could be made at the entrance, exit or both.
x On the other hand, Open Systems present just one barrier at the entrance of the freeway.
x Finally, Open Road Systems do not have barriers and payments are received by electronic collection14.
Thus, considering these three types of roads, each country has developed its own systems for the toll collection.
In particular, it is possible to identify two collection methods. Payments by cash, credit cards or special tickets are
the most common6. These collection systems are still present almost everywhere, but many developed countries are
using electronic toll collection. This method presents remarkable advantages, such as costs and congestion
reduction, no need for the speed decrease and consequently, no large delays. However, many countries are still using
both methods, without achieving total benefits of implementing electronic toll collection. In particular, many
European countries use different toll collection techniques, regardless of expectations that a unified system should
be used. Thus, this creates a fragmented global European transportation network15, in which each nation has its own
rules and payment systems. Presently, the European Union (EU) is developing a common transportation network,
Bruno Carlo Rampinelli Rota and Milan Simic / Procedia Computer Science 96 (2016) 1637 – 1646 1639
called European Electronic Toll Service (EETS), based on electronic toll collection. The main objective is to
develop and implement an efficient and ‘without borders’ system16.
2.1. Open Toll Road (ORT) design, approaches and operations
Open Toll Road (ORT) or Free-Flow Tolling refer to toll roads considered already as parts of ITS. The names
define exactly the types of tool collection scenarios on the roads. On one hand, the name ORT points out the fact
that these roads are constructed without barriers14. On the other hand, Free-Flow Tolling stresses on the importance
that vehicles pass through the payment without decreasing their speed, thus at the free-flow speed17. The free flow
speed is defined by government regulations as the maximum speed allowed. This key fact brings the following
benefits18: congestion minimization, time saving for drivers, cost saving for governments or freeways’ managing
companies and gas pollution diminution. We have investigated congestion minimization and time saving for the
combined tolling systems. Regarding to costs savings, recent studies show that OTRs result in significant cost
reduction. Construction costs are decreased due to the fact that toll bridge infrastructures replace common
tollbooths. Figure 1 shows the difference between common tollbooths (a) and electronic toll points (b). Moreover,
operations and maintenance costs are reduced because the payment systems are completely computerized and
wireless data acquisition (DAQ) is performed. Finally, focusing on the economical point of view, nations would
benefit from implementing a unique and standard payment toll system16. Further, according to other studies 19, 20,
OTRs improve air quality and promote ‘green transportations systems’21, which are strictly related to ITS. In
particular, researches have focused on carbon monoxide (CO) levels and they conclude that, in theory, it will be
possible to reduce CO emission up to 97%20. Thus, it is clear that OTRs brings important advantages in multiple
spheres of the transportations system and environment protection.
Fig. 1. (a) Common tollbooth (www.nj.com); (b) Electronic toll bridge (http://www.brisainovacao.pt)
2.2. Electronic Toll Collection (ETC), New Technologies and Future Applications
Electronic Toll Collection (ETC) systems have started developing since William Vickrey, Nobel Economic Prize
winner, speculated the use of electronic payment systems in order to avoid traffic congestions in the Washington
D.C. area22. The first technology adopted, required the users to slow down, up to a minimum limit speed, so that
infrastructure i.e. data acquisition system could identify passing vehicle. When the system detects and
communicates with the vehicles, it automatically opens the bar, allowing the vehicle to pass through the tollbooths.
In the most recent systems there are no such limitations and vehicles do not slow down for tooling. This is the case
of Open Road Systems, which allow a constant free flow, resulting in all the benefits mentioned before. Presently,
two systems for Automated Vehicle Identification (AVI) are used worldwide. The first technology consists of two
systems – an antenna at the toll point and a transponder installed on vehicles – which communicate using Dedicated
Short Range Communications (DSRC). The transponder can be referred to as an On Board Unit (OBU)23. These two
systems typically connect through the selected radio frequency (RF) channel, resulting in a reliable and efficient
system operation. The transponder and the antenna are part of the transportation network and they represent a
moving node – the transponder – and a fixed node – the antenna – of VANET3. While economic and environmental
advantages are proved, this system also presents two disadvantages: a cost increase for freeways’ users and
difficulties in identifying toll evaders. Drivers have to pay and install transponders at their own expense and this
could result in an effectiveness deficiency when governments or municipalities decide to switch from old non-
electronic system to ETC. Furthermore, the majority of transponders, Radio Frequency Identification (RFID) tags,
can be transferred from one vehicle to another with difficulties in recognizing toll evaders. In some countries,
transferring those electronics tags (etags) is allowed, as long as they are registered to more than one vehicle. The
best practice is to have a dedicated etag per each vehicle.
The second ETC system, the Automatic Number Plate Recognition (ANPR), does not create a communication
between the moving node and the fixed infrastructure. Using cameras, it reads vehicles plates and then addresses the
toll payment directly to the user (i.e. the vehicle’s owner). The system includes a series of cameras that detect plate
numbers through optical character recognition (OCR). After this first step, the system identifies the vehicle’s owner
and send him/her the payment bill24. This ETC technology brings multiple advantages: users’ costs reduction, users’
willingness to adopt this system and reliable identification of vehicles’ types. In fact, users do not have to buy a
compulsory device, resulting in a positive user behavior. However, studies24, 25 conclude that APNR has high
operational costs that do not justify its use. In the tooling systems that relay on RFID identification, as the most
reliable, image recognition is used as a backup system only when passing vehicle does not have tag installed. In that
case charges to the car owner are higher.
All those new technologies are primarily trying to optimize the actual services in order to improve the traffic
throughput and tooling reliability26. Considering the popularity of smartphones, IT companies have developed
mobile applications that directly communicate with tooling system antenna, providing the user with real-time
information on toll payments through push notifications. Theoretically, this technology could solve problems related
with ETC, offering high flexibility to users, which do not have to buy an OBU, and reliability, associated with low
cost, for roads’ managing companies27. This might be one of the optima solutions. Mobile phone applications have
already entered into many engineering, science, medical and other domains.
3. Traffic Flow: Theory and Algorithms
Three methods of traffic analysis, based on the level of scale and complexity, have gained importance from the
1950s. The microscopic scale relies on fluid dynamics models. The macroscopic scale is based on partial differential
equations (PDE). Kinetic scale expresses traffic theories using probability and statistics. Our research focused on the
microscopic model, since it brings particular advantages in describing traffic jam and bottlenecks.
Most important traffic flow theories are based on three parameters28, 29: flow (q), density (ȡ) and speed (v). Flow
defines the number of vehicles passing through a particular point in the unit of time [No/s]. Density is defined as the
number of vehicles in the unit of length [No/Km]. Speed refers to the mean speed of the traffic flow measured in
[m/s] and it is function of the density28. Theory formulation started in 1950 when M.J. Lighthill, G.B. Whitham and
P.I. Richard described their theory using fluid dynamics flow rules and laws, and applying them to traffic studies30.
In particular, they stated that traffic flow is a conserved process. Applying a fluid flow analogy, they wrote the non-
linear scalar conservation law for traffic, as shown in equation (1):
(1)
In this formula it is important to notice that the flow q(ȡ) is function of the density, ȡ, which also is function of
the time (t) and space (x). This formula represents the basis for any further study. Moreover, the inverse of the flow
(1/q) describes the space-time between the i-th vehicle and the (i+1)th vehicles that pass through a reference point.
This parameter is called headway and it is particularly useful in representing and modelling traffic congestions17.
Using the conservation law of equation (1), it is possible to refer it to two states, which in traffic models are two
arbitrary positions (x1 and x2), as expressed in equation (2).
(2)
This expression is useful for the aim of our investigation. In fact, the left side represents the increase in the
number of cars and the right side the flux before and after two points, which represent an arbitrary points before the
tollbooths and after the payment. The following equation (3) rewrites the previous equation as a PDE:
(3)
As said before, speed depends on density and equation (4) describes this relation:
(4)
In equation (4) vff is the free flow speed and ȡmax represents the critical density28. It means that if ȡ < ȡmax, there
is no congestions and if ȡ ȡmax, the bottleneck is forming jams. Thus, considering an initial density ( ), the traffic
problem has this system of equations:
(5)
Differentiating for t, the problem (5) leads to the solution expresses in equation (6):
(6)
, known as signal speed1, represents the wave front velocity at which shockwaves propagate their effects on
following vehicles.
In order to solve these equations, cumulative vehicles count curves, sometime referred as N curves, can be used.
They describe the totality of vehicles in the system that want to go over the toll payments in a certain time (i.e. the
time window), and the real number of users that are able to do so31. Three curves - arrival, virtual and departure
curve – define the traffic situation at bottlenecks in N curves. The arrival curve shows the vehicles at time T, the
virtual curve represents the number of vehicles that do not experience delay and the departure curve shows the real
traffic throughput considering the queue that it may form at the bottleneck. N curves allow to conduct a
mathematical model that requires these assumptions: constant deceleration for all vehicles, constant payment times,
no human errors or system failures. Thus, despite some human behaviour, such as time reaction32, and a
differentiation between cars and heavy vehicles are considered, other factors, such as drivers’ age and experience are
neglected. Considering that in 2017 UGVs will appear on highways33, these assumptions explore their positive
effects on transportation networks. In fact, UGVs can overall travel at constant speed, reducing variations and
deviations from mean value for parameters, such acceleration and deceleration.
4. The Study Case: BREBEMI
This paragraph introduces the case study chosen, explaining the freeway’s layout, showing 2015 data and
forecasts for the next ten years.
4.1. Real Life Example: Key Facts and Data
BreBeMi (Brescia-Bergamo-Milano), also known as A35, is an Italian highway that connects the city of Brescia
with Milan, passing through the Bergamo province, in north Italy34. It presents three lanes in each direction with a
regular flow of 40,000 vehicles/day. BreBeMi was the first Italian freeway realized in project financing, without
Government and State contributions, and it cost totally 1.61 billion of euros35.
BreBeMi S.p.A.34 provided data collected in 2015 and it predicted a five percent vehicle increase in the next 10
years. Table 1 shows data and forecast results for years 2016, 2019, 2022 and 2026. According to the data analyzed,
the tollbooth in Liscate, the closest one to Milan, is the freeway’s bottleneck. The tollbooth experiences two high-
demanding times (i.e. peak hours) – 7:30-9:30 am and 4:30-6:30 pm – that represent the time-window for the
analysis.
Table 1. Data and forecast for the BreBeMi freeway

Daily flow 2016 2019 2022 2026
Total number of vehicles 40,000 46,305 53,604 65,156
Total number of vehicles in Liscate in the pick hours 6,000 6,946 8,041 9,773
5. Traffic Throughput Maximization
This study analyses three cases in order to show throughputs. The mathematical model considers many
parameters to create a correct and reliable simulation. In particular, it was necessary to understand constraints for all
three models, formalizing criteria and constants, and defining decision variables. In order to solve the problem, IBM
ILOG CPLEX Optimization is used. This software requires36 data and variables declaration, objective functions and
constraints. Maximizing the throughput is the objective function. Table 2 shows all the objects, parameters,
constants, variables and data used.
Table 2. Mathematical model’s input and outputs

Decision variables and Data Nomenclature Value (cars/trucks) Unit
Total number of vehicles with Telepass N’ 4,200 in 2016 -
Total number of vehicles without Telepass N’’ 1,800 in 2016 -
Time window T 7200 s
Number of lanes (Telepass/Cash) L (Lp/Lc) - -
Free Flow speed vff 130/110 Km/h
Maximum speed in the payment lanes vpay 30 Km/h
Time between vehicles TBV - s
Time for make a non electronic payment Tpay 10 s
Boolean variables yi 1 if the ith vehicle passes; 0 otherwise -
Time at which each vehicle passes through the Ti To be found s
tollbooth
Throughput Z To be found No. of vehicles
5.1. The Mathematical Model
The mathematical model, used to describe the traffic throughput, implements the data in an iterative algorithm
that stops when it reach the optimal solution. The model states to find the maximum throughput, as equation (7)
shows:
(7)
The time at which each vehicle passes through the tollbooth is an important parameter because it describes the
efficiency of the system and sets the upper bound time constraint. Equation (8) defines Ti as a decision variable of
the algorithm. Naturally, Ti depends on TBV (time between vehicles), which is expressed in equation (9) and it is
function of the traffic flow. In equation (9), constants are all data such as mean time for payments, mean
deceleration time, traffic flow and density during peak hours that we received from BreBeMi S.p.A. However, it is
important to state that these values can be modified and adapted to other real life examples32.
(8)
(9)
Finally, inequality (10) sets the time limit and the algorithm is forced to stop when the time window is reached:
(10)
5.1.1. Case A: The Present Situation
The tollbooth in Liscate presents 6 lanes; four are dedicated to Telepass and 2 to cash payments. The technology
of Telepass works only in closed roads with barriers both at entrance and exit. Moreover, Telepass lanes allow a
maximum vehicle’s speed of 30 km/h for safety reasons37. Finally, cash lanes represent the real critical bottlenecks
because vehicles must stop for payment and have to wait for receipts, or change. These procedures form shockwaves
that result in traffic congestions. Thus, the mathematical model for vehicles with Telepass should consider times to
reduce speed, times to pass through the tollbooths and any delays. Vehicles that pay by cash should add the time to
stop and pay, which results in more critical shockwaves.
5.1.2. Case B: Four Telepass Lane
This model presents fewer constraints than Case A because it proposes to replace the cash lanes with other two
Telepass Lane. This solution involves that all users have a Telepass system installed. However, this proposal would
only be feasible if Italy adopted a unique toll payment system or if Europe imposed to its State Members the
introduction of EETS 16. Thus, this mathematical model considers the slowing down time from the free flow speed
to 30 km/h as the main constraint.
5.1.3. Case C: An Open Road Toll Design
The last model suggests the introduction of tollbooths without barriers in order to avoid speed reduction and
travel at the free flow velocity. This design involves the same complications discussed for case B. Thus, this
mathematical model does not present speed constraints – except for the maximum free flow speed allowed – and, in
order to show the ORT benefits, the algorithm considers only 3 lanes, instead of the six existing.
5.2. Results and Comparisons
CPLEX solves the algorithms for the three cases in respectively 98, 75 and 48 seconds for the year 2016 with
small timing increments for 2019, 2022 and 2026. Table 3 and figures 3,4 and 5 show the differences of the three
models in terms of vehicles’ throughput. In particular, it is relevant to point out that Case A is not able to satisfy the
demand because of the bottleneck created in the cash lanes that continuously expand traffic congestions during peak
hours.
Fig.3. Throughput results

Table 3. Results
Vehicles’ throughput 2016 (6,000 veh/2hrs) 2019 (6,946/2hrs) 2022 (8,041/2hrs) 2026 (9,772/2hrs)
Model A 5,864 6,546 7,312 8,500
Model B and C 6,000 6,946 8,041 9,772
Table 3 and figure 3 show that model B and C have outstanding results in terms of vehicles’ throughput and both
of them can achieve the maximum vehicles’ capacity required for each year, with a respectively maximum of 15,604
and 39,768 vehicles during peak hours. In particular, considering 2026, which is the year with the highest forecast,
model B is able to almost double the maximum capacity and model C can achieve four times the throughput
required. Thus, it is important to notice that model B suggests that introducing a unique electronic collection system
has particular benefits. The mathematical model of case C implies that ORTs represent the most performing systems
and these types of transportation network have the best result and output in terms of vehicles’ throughput and
environmental issues. Moreover, the result obtained shows that BreBeMi was correctly designed and constructed6.
Since the freeway presents the same characteristics – number of lanes, no tunnels and constant slope – for all of its
length, it is possible to conclude that BreBeMi could achieve important results switching from the present toll
collection system to ETC. On the other hand, figure 4 shows limitations of the present system. Focusing of 2026, it
is possible to observe that more than 1,200 vehicles, which represent the 13% of the vehicles during peak hours,
cannot pass through the tollbooth in the time window. Moreover, the mathematical model underlines that the
presence of cash lanes creates the tollbooth bottleneck.
Fig. 4. Model A bottleneck
6. Conclusion and Future Work
We have presented different toll payment methods, showing benefits and disadvantages that each technology brings
into the transportation system. The study presented in this paper shows the importance of communications and the
need to integrate different technologies and procedures together in order to develop ITSs. Using a real life example,
we investigated the benefits associated with ETC in term of vehicles’ throughput on freeways. ETC technology is
already available but it still carries problems and issues that limit its applications and outbreak. Despite that the
benefits of introducing ORT systems are evident and undeniable, further research in the technology implementation
should be conducted. In particular, future studies should focus on technologies that avoid using OBUs and allow an
absolutely reliable detection of vehicles, in order to avoid toll evaders and failures in the system. Finally, we
highlight the importance and the necessity to adopt a common toll payment system in the whole Europe. This
investigation strongly supports the idea of the European Electronic Toll Service (EETS) and it has shown economic,
environmental and social advantages in establishing this solution in EU. Further analysis should clearly state and
demonstrate that EU nations, with the aid of European public funds, would have commercial and economic benefits,
which could ensure a self-financing and a short payback period for this technology widespread implementation.
References
1. Davis, L.C., Predicting travel time to limit congestion at a highway bottleneck. Physica A: Statistical
Mechanics and its Applications, 2010. 389(17): p. 3588-3599.
2. Cheng, H., Autonomous Intelligent Vehicles Theory, Algorithms, and Implementation. Autonomous
intelligent vehicles: theory, algorithms, and implementation. 2011, London: Springer.
3. Simic, M., Vehicular ad hoc networks, B.D. Milovanovic, Editor. 2013, IEEE and University of Nis
(Serbia).
4. Campolo, C., A. Molinaro, and R. Scopigno, Vehicular ad hoc Networks Standards, Solutions, and
Research, ed. A. Molinaro and R. Scopigno. 2015, Cham: Springer International Publishing.
5. Kozlov, V.V., A.P. Buslaev, and A.S. Bugaev, Traffic and Granular Flow '11, ed. A.P. Buslaev and A.S.
Bugaev. 2014, Berlin: Springer Berlin Heidelberg.
6. Roess, R.P. and E.S. Prassas, Highway Capacity Manual A Conceptual and Research History. Highway
Capacity Manual, ed. E.S. Prassas. 2014, Cham: Springer.
7. Mazaré, P.-E., et al., Analytical and grid-free solutions to the Lighthill–Whitham–Richards traffic flow
model. Transportation Research Part B, 2011. 45(10): p. 1727-1748.
8. Kerner, B.S., The Physics of Traffic. Physics of Traffic. 2004: Springer Berlin Heidelberg.
9. Miller, A., A queueing model for road traffic flow. Journal of the Royal Statistical Society, Series B,
Methodological, 1961. 23(1): p. 64.
10. Ozguner, U., T. Acarman, and K. Redmill, Autonomous Ground Vehicles, ed. T. Acarman and K. Redmill.
2011, Norwood: Artech House.
11. Ratti, C.a.C., M., The City of Tomorrow: Sensors, Networks, Hackers, and the Future of Urban Life Yale
University Press, 2016.
12. Litman, T., Autonomous Vehicle Implementation Predictions. Victoria Transport Policy Institute, 2014. 28.
13. Singh, S.K., R. Duvvuru, and S.S. Thakur, Congestion Control Technique Using Intelligent Traffic and
Vanet. 2015.
14. Poole, R.W., Jr., Roads in a Market Economy. p. 67.
15. Report, M., Regional European Electronic Toll Service (REETS TEN). 2015: London.
16. Dionori, F., et al., Technology options for the European electronic toll service. 2015, Brussels: Brussels :
European Parliament.
17. Treiber, M., A. Kesting, and C. Thiemann, Traffic Flow Dynamics Data, Models and Simulation. Traffic
Flow Dynamics, ed. A. Kesting and C. Thiemann. 2012, Dordrecht: Springer.
18. Saravanan, S., A. Sabari, and M. Geetha, Fuzzy-based approach to predict accident risk on road network.
International Journal of Emerging Technology and Advanced Engineering, 2014. 4(5): p. 536-540.
19. Klodzinski, J., E. Gordin, and H. Al-Deek, Evaluation of impacts of open road tolling on main-line toll
plaza. Transportation Research Record: Journal of the Transportation Research Board, 2008.
20. Coelho, M., T. Farias, and N. Rouphail, Measuring and modeling emission effects for toll facilities.
Transportation Research Record: Journal of the Transportation Research Board, 2005(1941): p. 136-144.
21. Xiao, Y. and A. Konak. Green Vehicle Routing Problem with Time-Varying Traffic Congestion. in
Proceedings of the 14th INFORMS Computing Society Conference. 2015.
22. Kelly, F., Road pricing: addressing congestion, pollution and the financing of britain's roads. Ignenia,
2006. 29(1): p. 34-40.
23. Simic, M., Contemporary digital radio technologies, ed. M. Simic. 2006, Melbourne: RMIT University.
24. Parisi, R., et al. Car plate recognition by neural networks and image processing. in Circuits and Systems,
1998. ISCAS'98. Proceedings of the 1998 IEEE International Symposium on. 1998. IEEE.
25. Lotufo, R., A. Morgan, and A. Johnson. Automatic number-plate recognition. in Image Analysis for
Transport Applications, IEE Colloquium on. 1990. IET.
26. Thomopoulos, N., M. Givoni, and P. Rietveld, ICT for transport : opportunities and threats, ed. N.e.
Thomopoulos, M.e. Givoni, and P.e. Rietveld. 2015: Edward Elgar Publishing.
27. Aalsalem, M.Y. and W.Z. Khan, CampusSense-A Smart Vehicle Parking Monitoring and Management
System using ANPR Cameras and Android Phones. Journal Editorial Board, 2016: p. 809.
28. Perlman, C., Mathematical modelling of traffic flow at bottlenecks. 2008, Master’s Thesis. Lund Institute of
Technology. Advisor: Stefan Diehl.
29. López Vidal, A., Traffic flow simulation and optimization using evolutionary strategies. 2011.
30. Lighthill, M.J. and G.B. Whitham. On kinematic waves. II. A theory of traffic flow on long crowded roads.
in Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences. 1955.
The Royal Society.
31. Cassidy, M.J. and R.L. Bertini, Observations at a freeway bottleneck. Transportation and Traffic Theory,
1999: p. 107-146.
32. Maurya, A.K. and P.S. Bokare, Study of deceleration behaviour of different vehicle types. International
Journal for Traffic and Transport Engineering, 2012. 2(3): p. 253-270.
33. Munsell, D.T., Praveen; Hogarth, James, The New Road to the Future: Realising the benefits of
Autonomous vehicles in Australia. Accenture Digital, 2014.
34. SpA, S.d.P.B., Motorway Link Connection between Brescia and Milano. Infrastructural Lombard System,
2014.
35. Lusini, L., Lo strumento del project financing nel settore dei trasporti: il caso Brebemi. 2014.
36. Hillier, F.S.a.L., Gerald J., Introduction to Operations Research. 2001.
37. Greco, A. and G. Ragazzi, History and regulation of Italian highways concessionaires. Research in
Transportation Economics, 2005. 15(1): p. 121-133.
Nama : Kayla Dwi Anindya
NIM : 3336220124
Metode Penelitian (B)
Tabel 1.1 Hasil Review Jurnal

No Judul Penelitian Rumusan Masalah Tujuan Penelitian Metode Hasil
1. Penerapan Manajemen 1. Bagaimana kinerja Penelitian ini Metode penelitian Studi ini menunjukkan
Lalu Lintas Satu Arah ruas jalan dan simpang bertujuan untuk ini meliputi tahap bahwa penerapan
pada Ruas Jalan Sultan serta mencari solusi mengidentifikasi persiapan dengan manajemen lalu lintas satu
Agung Kota Semarang penyebaran lalu lintas skenario terbaik studi kepustakaan arah pada ruas jalan di
untuk Sebaran Beban yang efisien dengan dalam penerapan lalu terkait manajemen Kota Semarang dapat
Lalu Lintas. merencanakan dua manajemen lintas satu lalu lintas satu arah meningkatkan kinerja lalu
skenario manajemen arah di Kota dan kondisi lalu lintas dan mengurangi
lalu lintas sistem satu Semarang guna lintas di Kota tundaan kendaraan. Dari
arah? meningkatkan Semarang. dua skenario yang
2. Apakah penerapan kelancaran dan Selanjutnya, dibandingkan, skenario 1
efisiensi lalu lintas. dilakukan tanpa contra flow terbukti
sistem satu arah pada
Penelitian ini juga pengumpulan data lebih efektif daripada
ruas jalan dan simpang
bertujuan untuk melalui survei skenario 2 dengan contra
dapat mengurangi menyalakan kinerja lapangan untuk flow. Skenario 1 berhasil
tundaan lalu lintas? ruas jalan dan mendapatkan menurunkan nilai DS pada
3. Apakah skenario 1 simpang serta informasi volume lima ruas jalan dan tidak
tanpa contra flow mencari solusi kendaraan, ada waktu tundaan,
memberikan hasil penyebaran lalu lintas geometrik jalan, sementara skenario 2
terbaik dalam yang efisien dalam dan kondisi hanya berhasil
mengatasi kemacetan rangka mengatasi lingkungan. Data menurunkan nilai DS pada
lalu lintas di Kota hamparan kendaraan kemudian tiga ruas jalan. Dengan
Semarang? di Kota Semarang. dianalisis menggunakan metode
Selain itu, tujuan berdasarkan teori scoring
penelitian ini adalah kapasitas dan
untuk memberikan kinerja lalu lintas berdasarkan kecepatan
rekomendasi terkait pada MKJI 1997. rencana, skenario 1
penerapan sistem satu Penataan lalu memperoleh hasil tertinggi
arah pada ruas jalan lintas dilakukan dalam evaluasi kinerja lalu
tertentu guna dengan membuat lintas. Rekomendasi yang
meningkatkan kinerja dua skenario diberikan meliputi
lalu lintas di kota rencana yang pemilihan rute yang minim
tersebut. dibandingkan konflik, penerapan
dengan kondisi skenario 1 pada ruas
eksistingnya. Hasil tertentu, analisis lebih
analisis kinerja lanjut, dan pemberian
ruas jalan dan marka dan rambu lalu
simpang lintas yang jelas. Penelitian
dievaluasi ini memberikan kontribusi
menggunakan penting dalam
metode scoring meningkatkan kelancaran
untuk menentukan dan efisiensi lalu lintas di
skenario terbaik. Kota Semarang melalui
Penelitian ini penerapan sistem satu
menggabungkan arah pada ruas jalan
analisis lapangan, tertentu.
teori kapasitas
jalan, dan evaluasi
skenario untuk
mencapai tujuan
penelitian.
2. Analisa Lalu Lintas Bagaimana dampak Tujuan dari penelitian Penelitian ini Studi ini membahas
Penerapan Sistem Satu penerapan Sistem Satu ini adalah untuk menggunakan dampak penerapan
Arah di Kawasan Arah (SSA) di kawasan mengidentifikasi metode Sistem Satu Arah (SSA)
Dukuh Atas, Jakarta Dukuh Atas, Jakarta pengaruh penerapan mikrosimulasi terhadap lalu lintas di
terhadap kinerja lalu Sistem Satu Arah dengan perangkat Kawasan Dukuh Atas,
lintas, termasuk (SSA) terhadap lunak Vissim 11
peningkatan kecepatan kecepatan lalu lintas untuk Jakarta. Hasil studi
rata-rata dan dan tundaan di memperkirakan menunjukkan bahwa
pengurangan tundaan simpang-simpang di dampak lalu lintas penerapan SSA
di simpang? Apakah Kawasan Dukuh penerapan Sistem menghasilkan
rekomendasi dapat Atas, Jakarta. Selain Satu Arah (SSA) peningkatan signifikan
diberikan untuk itu, penelitian ini juga di Kawasan Dukuh dalam kecepatan rata-rata
menerapkan SSA di bertujuan untuk Atas, Jakarta. lalu lintas dan
bagian barat kawasan memberikan Tahapan penelitian
pengurangan tundaan di
Dukuh Atas? rekomendasi terkait meliputi
simpang-simpang di
penerapan SSA di pengumpulan data
kawasan Dukuh Atas volume lalu lintas, kawasan tersebut.
bagian barat. pemodelan lalu Kecepatan rata-rata lalu
lintas dengan lintas meningkat dari 15,5
pendekatan km/jam menjadi 17,7
pembebanan lalu km/jam, sementara
lintas statis, tundaan pada simpang
kalibrasi dan berkurang dari 60,3
validasi model, detik/kendaraan menjadi
serta analisis 43,7 detik/kendaraan
perbandingan setelah penerapan SSA.
antara kondisi
eksisting dan
kondisi rencana
SSA. Selain itu,
survei
inventarisasi jalan
dilakukan untuk
mendata ruas
jalan, simpang,
tipe jalan, lebar
jalan, dan
pengaturan lalu
lintas. Hasil output
model eksisting
dari Vissim
divalidasi dengan
data survei
lapangan untuk
memastikan
konsistensi model
dengan hasil
lapangan
3. Analisis Transportasi 1. Bagaimana tingkat Tujuan penelitian ini Metode penelitian Hasil penelitian
Terhadap Penerapan pelayanan jalan Kol. adalah untuk ini meliputi studi menunjukkan bahwa
Arus Lalu Lintas Satu Amir Hamzah yang menganalisis dampak literatur untuk penerapan arus lalu lintas
Arah merupakan jalan satu penerapan arus lalu mengumpulkan satu arah di Kota Jambi
arah dengan kondisi lintas satu arah di teori-teori dan memiliki dampak
perkerasan yang Kota Jambi terhadap rumus-rumus signifikan terhadap
kurang baik dapat pengguna jalan dan terkait, serta pengguna jalan dan kinerja
ditingkatkan melalui kinerja jalan. Selain pengumpulan data jalan. Studi ini
pengaturan parkir yang itu, penelitian juga primer dan mengungkapkan bahwa
lebih baik, penyediaan bertujuan untuk sekunder untuk volume lalu lintas dan
mengamati volume analisis volume hambatan samping di jalan
fasilitas parkir yang
lalu lintas dan lalu lintas dan Kol. Amir Hamzah
memadai, dan
hambatan samping di hambatan samping merupakan jalan satu arah
pelebaran jalan jika jalan Kol. Amir di jalan Kol. Amir dengan kondisi perkerasan
memungkinkan? Hamzah, serta untuk Hamzah. yang kurang baik.
mengevaluasi kondisi Pengumpulan data Kecepatan arus lalu lintas
kinerja jalan, hambatan samping dan kapasitas ruas jalan
kecepatan arus lalu dilakukan dengan juga telah dihitung dalam
lintas, dan kapasitas menghitung dan penelitian ini. Saran
ruas jalan tersebut. mencatat aktivitas perbaikan yang diajukan
Saran-saran samping jalan untuk meningkatkan
perbaikan juga selama periode tingkat pelayanan jalan
ditujukan untuk pengamatan yang termasuk pengaturan
meningkatkan tingkat dilakukan oleh parkir yang lebih baik,
pelayanan jalan, surveyor. Analisis persyaratan parkir bagi
termasuk dalam hal kinerja dan tingkat pemilik bangunan, Dengan
pengaturan parkir pelayanan jalan demikian, penelitian ini
yang lebih baik, dilakukan dengan memberikan kontribusi
persyaratan parkir menghitung penting dalam pemahaman
bagi pemilik derajat kejenuhan tentang transportasi dan
bangunan, dan (DS) berdasarkan rekayasa lalu lintas di Kota
pelebaran jalan jika arus total Jambi, serta memberikan
memungkinkan. sesungguhnya dan rekomendasi untuk
kapasitas perbaikan infrastruktur
sesungguhnya. jalan yang dapat
Metode penelitian meningkatkan efisiensi
ini didasarkan dan keamanan lalu lintas di
pada standar dan wilayah tersebut.
panduan rekayasa
lalu lintas, seperti
Manual Kapasitas
Jalan Indonesia
(MKJI).
4. Penerapan Sistem Lalu Bagaimana kinerja arus Tujuan penelitian ini Metode penelitian Hasil penelitian
Lintas Satu Arah pada lalu lintas di Simpang adalah untuk ini melibatkan menunjukkan bahwa
Simpang Tiga Straat A Tiga Straat A dan ruas menganalisis kinerja survei lapangan penerapan sistem lalu
di Kota Kupang: jalan terkait di Kota sistem lalu lintas satu selama 6 hari di 4 lintas satu arah pada
Analisis Kinerja dan Kupang, serta untuk arah yang diterapkan titik lokasi simpang tiga Straat A di
Rekomendasi mengevaluasi dampak pada simpang tiga penelitian di Kota Kota Kupang telah
Penyempurnaan penerapan sistem lalu Straat A di Kota Kupang, dengan meningkatkan kinerja
lintas satu arah Kupang, serta pengumpulan data simpang tersebut dengan
terhadap kinerja memberikan primer berupa tingkat pelayanan yang
simpang tersebut? rekomendasi untuk kondisi geometri, lebih baik. Volume lalu
penyempurnaan yang volume lalu lintas, lintas mengalami
dapat meningkatkan dan hambatan penurunan, hambatan
kelancaran lalu lintas samping, serta data samping tergolong sedang,
dan mengurangi sekunder berupa kecepatan kendaraan
tundaan serta peluang penelitian meningkat, dan tingkat
antrian di lokasi terdahulu dan data kejenuhan yang lebih
tersebut. jumlah penduduk. rendah setelah penerapan
Penelitian ini sistem satu arah.
menggunakan Rekomendasi yang
metode analisis diberikan meliputi
data survei untuk pemberhentian sementara
mengevaluasi bagi mobil penumpang,
kinerja sistem lalu penyesuaian kelas jalan
lintas satu arah dan geometri untuk
pada simpang tiga meningkatkan kapasitas
Straat A di Kota dan keselamatan jalan.
Kupang. Studi ini memberikan
kontribusi penting dalam
upaya meningkatkan
kelancaran lalu lintas di
Kota Kupang yang sedang
mengalami perkembangan
pesat. Referensi manual
kapasitas jalan Indonesia
(MKJI) dan Highway
Capacity Manual (TRB
2000) digunakan dalam
analisis untuk mendukung
kesimpulan tersebut.
5. Analisis Kinerja Bagaimana kinerja Tujuan dari penelitian Penelitian ini Berdasarkan analisis data
Simpang Bersinyal dan simpang bersinyal dan ini adalah untuk menggunakan lalu lintas, waktu tempuh,
Ruas Jalan Setelah ruas jalan setelah menganalisis kinerja metode observasi dan tingkat kepadatan
Diterapkan Sistem Satu diterapkan sistem satu simpang bersinyal lapangan dan kendaraan setelah
Arah pada Jalan Dewi arah pada Jalan Dewi dan ruas jalan setelah analisis data untuk penerapan sistem satu arah
Sartika di Depok Sartika di Depok? diterapkan sistem satu mengumpulkan di jalan Dewi Sartika,
arah pada jalan Dewi informasi tentang Depok, ditemukan bahwa
Sartika di Depok. kinerja simpang sistem tersebut berhasil
Penelitian ini bersinyal dan ruas mengurangi kemacetan
bertujuan untuk jalan setelah lalu lintas dan
mengevaluasi diterapkan sistem meningkatkan kelancaran
efektivitas dari satu arah pada arus lalu lintas di wilayah
penerapan sistem satu jalan Dewi Sartika tersebut. Kinerja simpang
arah tersebut dalam di Depok. Data bersinyal dan ruas jalan
mengurangi lalu lintas, waktu juga menunjukkan
kemacetan lalu lintas tempuh, dan peningkatan setelah
dan meningkatkan tingkat kepadatan penerapan sistem satu
kelancaran arus lalu kendaraan arah. Rekomendasi
lintas di wilayah dikumpulkan perbaikan yang diberikan
tersebut. Selain itu, secara langsung di meliputi peningkatan
penelitian ini juga lapangan. Selain koordinasi sinyal lalu
bertujuan untuk itu, wawancara lintas, penambahan jalur
memberikan dengan pihak khusus transportasi umum,
rekomendasi terkait juga dan peningkatan kapasitas
perbaikan yang dapat dilakukan untuk jalan untuk
meningkatkan kinerja mendapatkan mengoptimalkan kinerja
simpang bersinyal informasi yang sistem satu arah tersebut
dan ruas jalan secara relevan. Data yang secara keseluruhan.
keseluruhan. terkumpul Dengan demikian, dapat
kemudian disimpulkan bahwa
dianalisis penerapan sistem satu arah
menggunakan pada
teknik statistik dan jalan Dewi Sartika di
perangkat lunak Depok memberikan
khusus untuk dampak positif terhadap
mengevaluasi
kinerja sistem satu kinerja lalu lintas di
arah tersebut. wilayah tersebut
6. Modeling traffic flow 1. How to reduce The research aims to The research The study presented in the
on two-lane roads with energy investigate the utilizes a multi- paper focuses on the
traffic lights and consumption and impact of drivers' agent system impact of drivers'
countdown timer emissions in road behaviors on traffic combined with a behaviors on traffic flow
transportation flow efficiency using cellular automaton efficiency in urban areas,
systems while countdown timers at model to simulate particularly concerning the
maintaining traffic lights, with a traffic flow use of countdown timers at
efficient traffic flow focus on improving dynamics on a traffic lights. A simulation
and mobility? traffic management two-lane road with system based on cellular
2. How to optimize strategies and traffic lights and automaton and multi-agent
countdown timers. technology was developed
traffic flow to reducing congestion
The model to model traffic flow on a
reduce congestion in urban areas.
incorporates two-lane road accurately.
and improve overall various driver The research aims to
system efficiency? behaviors, such as enhance traffic flow
3. How to utilize driver skills, efficiency, reduce
deterministic and carelessness, and congestion, and minimize
probabilistic reactions to environmental pollution by
models to estimate countdown timers, incorporating various
emissions and to analyze their driver behaviors into the
energy impact on traffic model. The proposed
consumption with flow efficiency approach combines a
physical meaning and congestion cellular automaton grid for
and Implement levels. road representation with
energy-efficient Experimental agents representing drivers
control strategies at studies are with specific behaviors.
the vehicle level by conducted to Factors such as driver
modifying speed evaluate the skills, carelessness, and
profiles and route effectiveness of reactions to countdown
choices? countdown timers timers at traffic lights are
in improving considered in the model to
traffic flow and analyze their influence on
reducing traffic flow and congestion
congestion. levels. Experimental
results demonstrate the
significant impact of these
factors.
7. Modeling traffic flow what impact of the The purpose of this The simple and This paper is focused on
on two-lane roads with traffic lights and their jurnal is to develop a effective applying computer
traffic lights and positive influence on multi-agent system computational simulation in modelling a
countdown timer traffic flow was also to study the capacity method of traffic traffic flow on two-lane
presented? of a road in relation was proposed by roads with traffic lights in
to the various Nagel and relation to different
behaviours of drivers Schreckenberg drivers’ behaviour. A
starting moving (Nagel & model was developed and
when the green light Schreckenberg, implemented in the form
is switched on. A 1992). The of a simulation system,
traffic simulator was authors which served to carry out
developed, on the formulated a the research study
basis of which stochastic model described herein. The
numerous of cellular basic CA traffic model
experimental studies automaton for has been modified in the
were carried out. single-lane roads context of the size of the
(NaSch model). In CA cell, which
the developed contributed to achieving
model time and lower vehicle speeds and
space are allowed for a more
discretized, and realistic mapping of
due to low vehicle traffic in the city.
computational In addition, various
complexity, the dependencies have been
model can be used introduced to symbolize
to model a large the different behaviour of
number of drivers who move when
vehicles. green light comes on.
Countdown timers have
been introduced to count
down the time until green
light turns on.
Experimental studies
were carried out and the
results are shown in the
figures. Positive
influence of countdown
timer has been
demonstrated and it has
been shown that
deficiencies in driving
skills and carelessness in
guarding the change of
lights translate into delays
in road traffic and thus for
greater congestion and
greater local
environmental pollution.
8. Ecological tra_c How road An overview of the This research uses This paper surveys the
management: a review transportation problems existing vehicle and quantitative existing emission and
of the directly addressing tra_c control methods where energy consumption
modeling and control the issue of energy strategies to improve the data is models, as well as the
strategies for improving consumption and obtained from traffic control strategies
environmental pollutant emissions energy and experimental and to reduce them, either by
sustainability of road reduction? environmental statistical tests. considering
transportation effeciency of vehicles independently,
transportation is or by considering traffic
given. The flows. The main
review focuses on advantages
the control and drawbacks of the
techniques that different approaches are
explicitly address highlighted.
energy consumption The first step to estimate
and emissions. The emissions and energy
connection and consumption is to
interaction between measure,
traffic or simulate, the
congestion and kinematics of vehicles,
energy effeciency is that can be either static or
also discussed. As an dynamic.
outcome of this Traffic models can be
review, research implemented on a
gaps in the current microscopic or
state of the art macroscopic scale.
have been identi_ed
and discussed in
order to inspire
future works in this
field.
9. A Brief Review of How to art of multilevel Introduction and Three main This paper contains a
Holonic Multi-Agent models for traffic and presentation of approaches are review of relevant papers
Models for Traffic and transportation modeling holonic principles, presented in in the field of holonic
and simulation using concepts and literature to model multi-agent systems and
framework, the road traffic. for traffic modeling and
Transportation Systems holonic multi-agent paper summarizes instance: simulation. Table 1
approach? research works on microscopic, summarizes the core
traffic models using intermediate characteristics of the cited
holonic multi-agent (mesoscopic and works, and highlights the
systems. Then a hybrid or similarities and
discussion is given, multilevel) and differences between the
in particular, a macroscopic approaches. In Section 4,
rationale of holonic approach. These some of the current
systems in modeling approaches problems facing with
and simulation of differentiate the HMAS are identified.
traffic and time and space HMAS research
transportation and scales. Multilevel metodology requires
open issues are approach further effort (in
given. combines the modeling of individuals)
advantages of and a consolidation of the
macroscopic level approaches pursued by
and microscopic the different research
level but it may be groups, especially those
difficult to realize. focus on the environment
and traffic control is
needed.
10. Traffic Flow How to define different This study analyses Three methods of This jurnal presented
Optimization on types of toll payment three cases in order traffic analysis, different toll payment
Freeways methods adopted in to show throughputs. based on the level methods, showing
freeways? The mathematical of scale and benefits and
model considers complexity, have disadvantages that each
many gained technology brings
parameters to create importance from into the transportation
a correct and reliable the system. The study
simulation. In 1950s. The presented in this paper
particular, it was microscopic scale
necessary to relies on fluid shows the importance of
understand dynamics models. communications and the
constraints for all The macroscopic need to integrate different
three models, scale is based on technologies and
formalizing criteria partial differential procedures together in
and constants, and equations (PDE). order to develope.
defining decision Kinetic scale Despite that the
variables. In order to expresses traffic benefits of introducing
solve the problem, theories using ORT systems are evident
IBM probability and and undeniable, further
ILOG CPLEX statistics. This research in the
Optimization is research focused technology
used. This software on the implementation
requires36 data and microscopic should be conducted. In
variables model, since it particular, future studies
declaration, brings particular should focus on
objective functions advantages in technologies that avoid
and describing traffic using OBUs and allow an
constraints. jam and absolutely reliable
Maximizing the bottlenecks. detection of vehicles, in
throughput is the order to avoid toll evaders
objective function. and failures in the system.
Table 2 shows all the Finally, we
objects, parameters, highlight the importance
constants, variables and the necessity to adopt
and data used. a common toll payment
system in the whole
Europe. This
investigation strongly
supports the idea of the
European Electronic Toll
Service (EETS) and it has
shown economic.
11. Dynamic one-way How two-way track This study aims to Numerical Addressed the issue of
traffic control in control is related to determine project on experiments show real-time control of a two-
automated junction control underground freight that our control way tube. Several new
transportation systems supported by traffic transportation using rules reduce control rules were defined
signals, see Newell Automated Guided waiting times by as an alternative to the
(1988) for an overview? Vehicles, single 10–25% classic periodic control
lanes for compared to rule. The adaptive
traffic in two a straightforward control rules try to use
directions are periodic rule. local information on
constructed to Dynamic queue lengths and look-
reduce infrastructure programming ahead information on
investment. yields the best future arrivals to improve
results in terms of on system performance.
mean waiting Simulation results
times. indicated that, depending
on the amount of
information available,
adaptive control rules
improve system
performance up
to 10% for high arrival
rates and up to 25% for
low arrival rates. In
general, the performance
of the rules was not much
influenced by the length
of the information
horizon. An experiment
that included the twoway
tube in a closed network,
the Schiphol system,
showed that our
conclusions still hold
for a closed system. An
interesting topic for
future research is a more
thorough examination
of the interaction effects
with terminals or other
two-way tubes within an
open or closed network.

Combine PDF

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Combine PDF

Uploaded by

Copyright:

Available Formats

Available online at www.sciencedirect.

Transportation Research Procedia 39 (2019) 300–308

Green Cities 2018

2352-1465  2019 The Authors. Published by Elsevier B.V.

4. The system developed for the simulation

Fig. 3. Class diagram.

Fig. 4. The main window of the developed road traffic simulator.

Fig. 6. The impact of drivers with little skill on traffic flow.

Fig. 7. The impact of carefree drivers on traffic flow.

Fig. 8. The impact of traffic lights countdown timer on traffic flow.

ANALISIS TRANSPORTASI TERHADAP PENERAPAN ARUS LALU LINTAS

PENDAHULUAN Dengan skala yang cukup besar, ternyata

58 Fakultas Teknik UMSB ISSN 2599-2081

ISSN 2599-2081 Fakultas Teknik UMSB 59

16.00- 12 161 32 Tabel 3. Lanjutan

60 Fakultas Teknik UMSB ISSN 2599-2081

Analisis Kinerja dan Tingkat Pelayanan Tabel 5. Tabel Perbandingan

ISSN 2599-2081 Fakultas Teknik UMSB 61

4. Derajat kejenuhan ( DS) atau tingkat

62 Fakultas Teknik UMSB ISSN 2599-2081

Dynamic one-way traﬃc control in automated

Keywords: Traﬃc control; Simulation; Transportation networks; Heuristics

Fig. 1. A two-way tube connecting double tube systems.

(a) Signiﬁcant throughput time in the two-way tube (several minutes).

(c) Embedding adaptive control rules in a rolling horizon environment.

3. Assumptions and control rule concepts

Our models are based on the following assumptions:

(1) All AGVs are identical.

3.2. Characterisation of control rules

4.1. Adaptive local control (ARloc)

4.1.1. The case d ¼ 0

4.1.2. The case d > 0

4.2. Adaptive look-ahead control (ARla)

4.2.1. The case d ¼ 0

4.2.2. The case d > 0

tkl ; T þ Cðt0 Þ þ d ðqr ðt0 þ Cðt0 ÞÞ 1Þg ð8Þ

þ 2T tkl þ d ðqr ðmaxftil ; t0 þ d ði 1Þg þ T Þ 1ÞÞ

4.3. Dynamic programming (DP)

4.3.1. The case d ¼ 0

fR ði; j; t Þ ¼ min fc½ði; j k2 ; t0 Þ; ði; j; t00 Þ þ fL ði; j k2 ; t0 Þg

4.3.2. The case d > 0

5. Design of the simulation study

6. Analysis of simulation results

6.1. Default settings

normalized average delay

6.2. Alternative system conﬁgurations

normalized average delay

normalized average delay

6.3. Sensitivity for the information horizon and planning frequency

7. A two-way tube in a closed system

7.1. The eﬀects of convoys on the system

7.2. Numerical results for the Schiphol-case

Fig. 6. Layout of the Schiphol system.

PENERAPAN MANAJEMEN LALU LINTAS SATU ARAH

Ramadhania Pramanasari, Nurul Qomariyah, Djoko Purwanto*), Epf. Eko Yulipriyono*)

Jurusan Teknik Sipil, Fakultas Teknik, Universitas Diponegoro

keywords: traffic density, one way system,traffic performance

Indonesia merupakan negara berkembang yang memiliki pertumbuhan penduduk yang

Semarang sebagai ibukota Provinsi Jawa Tengah dengan pertumbuhan penduduknya

Maksud dan Tujuan

Jl. Sultan Agung seksi

Jl. Sultan Agung seksi Jl. Dr. Wahidin

Jl. Teuku Umar

Gambar 1. Lokasi studi

fR ði; j; t Þ ¼ min fc½ði; j k2 ; t0 Þ; ði; j; t00 Þ þ fL ði; j k2 ; t0 Þg

Ramadhania Pramanasari, Nurul Qomariyah, Djoko Purwanto), Epf. Eko Yulipriyono)