You are on page 1of 8

Calibration of Conventional Macroscopic Traffic Flow Models for Nepalese

Roads:
(A Case Study of Jadibuti - Suryabinayak SectionN)
Hemant Tiwari1, Anil Marsani2
1
Graduate Student, Department of Transportation Engineering, Central Campus, Pulchowk, IOE
2
Coordinator, M.Sc. in Transportation Engineering, Department of Civil Engineering, Central Campus, Pulchowk, IOE
Corresponding Email: hemu.ioe@gmail.com

Abstract: The importance of understanding the relationship between speed, flow and density is unquestionable, from
the perspective of planning, design, traffic operations, and improvement of facilities. This particular research tries to
address the necessity of understanding the relationship between traffic flow parameters via calibration of various
macroscopic traffic flow models, which will provide quantitative estimates of the change in speed as a function of
anticipated changes in traffic demand for Jadibuti - Suryabinayak section. Various conventional macroscopic speed-
density models namely Greenshields, Greenberg, Underwood, Drake, Pipes – Munjal, Polynomial and modified
Drake and Underwood models, were first calibrated and further validated. Due to significant goodness of fit (R2 value
and standard error) and realistic estimates of free flow speed, optimum speed and optimum density for the section
under consideration, Conventional Underwood model was recommended as the best fit model. This calibrated model
can be used in predicting speed for the anticipated traffic volume, and even helps in quantifying congestion for
oversaturated condition.
Keywords: Macroscopic traffic flow model; PCU; free flow speed; jam density; goodness of fit

1. Introduction from the standpoint of highway planning, design and


operation as it provides quantitative estimates of the
Tremendous developments in the transportation change in speed as a function of anticipated changes in
system along with increased road network and traffic demand. These traffic flow models have a
overdependence on automobiles give rise to the significant role on prediction of highway capacity, free
evolution of various traffic flow theories. These flow speed and level of service offered by the facilities
theories illustrate traffic stream characteristics which and to determine the geometric features of road so as
includes behavior of vehicles on road at different to provide the targeted level of service. They are
circumstances, interaction of road user and roadway equally useful in real-time traffic control, a basic
environment. Researcher have categorized various component of Intelligent Transportation System (ITS)
traffic flow models under two categories viz. which is based on changes in traffic flow parameters.
macroscopic and microscopic. Conventional Calibration and validation of various macroscopic
mathematical models that formulate the relationships models is the first step towards creating theoretical
among the integrated parameter associated with traffic background for successful planning and design.
stream i.e. speed, flow and density are termed as
macroscopic traffic flow model. Such models convert
single entity level characteristics to comparable 2. Objectives of Study
system level characteristics by integrating microscopic
Despite being the widely accepted macroscopic traffic
traffic flow model, thus looking the traffic flow from
flow model, it is not universal that Greenshield model
the broader perspective. Greenshield, Greenberg,
will always be the best fit model. Many countries had
Underwood, Drake model etc. are some of the
calibrated macroscopic models for different parts or
examples of macroscopic traffic flow models.
even at different segment of roads, but our country
Microscopic traffic flow model on the other hand deals
lacks such calibration till date. Thus the objectives this
with the traffic stream characteristics from
research work is to identify conventional macroscopic
microscopic view point which includes headway,
traffic flow models; develop, calibrate and validate
clearance etc. via modeling driver-driver and driver-
them for heterogeneous traffic flow condition and
road interactions within a traffic stream. Car following
recommend the most suitable one.
model, Pipe’s model, Forbes’ model are some of the
microscopic traffic flow models.
The importance of understanding the relationship 3. Study Area
between speed, flow and density is unquestionable

1
Calibration of traffic flow models is of great the dominant mode is less than 80% of the traffic while
importance on urban areas where large traffic flow Fazio, Hoque, and Tiwari (1999) suggest the value to
occurs and thus Kathmandu area was chosen. Various be slightly higher at 85%. Only a few model are
constraints, such as time, manpower and fund c lead to develop in context of mixed traffic condition, some
confine the study area to Jadibuti – Suryabinayak road model convert heterogeneous traffic to equivalent
section, along the Araniko Highway. It serves passenger car-units and then applying procedures for
significant traffic from Kathmandu to Bhaktapur, homogeneous traffic.
forms a part of Asian Highway 42 running towards the
In case of heterogeneous traffic flow condition,
Chinese border with significant traffic volume and has
various types of vehicles need to be converted to
one of the highest traffic volumes among the trunk
equivalent vehicle using Passenger Car Unit (PCU) or
roads in Nepal. Moreover, it is anticipated that the road
Passenger Car Equivalent (PCE) factor. The PCU
will get much more congested than at present once the
value of a given type of a vehicle is related to its lane
Sindhuli Road is completed. At the same time this road
occupancy relative to a standard passenger car, whose
section is characterized by operating traffic signals
PCU value is taken as 1. NRS 2070 has provided PCU
which is rare to see in Kathmandu valley.
values for different types of vehicles. PCE values
depend on the traffic flow parameters and are
4. Literature Review subjected to variations due to the factors that influence
the traffic flow parameters. Therefore it may not be
In general, traffic streams are not uniform but vary precisely correct to adopt a constant set of PCE values
over both space and time, thus it is a difficult to under different roadway and traffic conditions (Justo
describe traffic flow phenomenon without the use of a and Tuladhar, 1984).
common set of terms: volume or flow rate, speed, and
density or concentration. Greenshield’s Model
Average speed (V) is defined as average spot speed of The first simplest classical macroscopic traffic flow
several vehicles passing a specific location. Flow or model was proposed by Greenshield’s in 1935 which
volume (Q) is defined as the number of vehicles assume that under uninterrupted flow conditions,
crossing a particular section within stated period of speed and density are linearly related. This model is
time. Volume and flow are interchangeably used in considered as the powerful tool in the field of traffic
some literature whereas in Highway Capacity Manual flow modeling due to its simplicity and reasonable
(HCM) flow rate corresponds to equivalent hourly rate goodness of fit but yet its acceptability is not universal
at which vehicles pass over a given point or section of as it don’t yield reasonable goodness of fit to data
a roadway during a given time interval of less than 1 corresponding to congested and extremely low density
hour, usually 15 minute. Density or concentration (K) road. Mathematically, the model is written as:
is defined as the number of vehicles occupying the 𝐊
particular section of road. 𝐔 = 𝐔𝐟 ∗ {𝟏 − }
𝐊𝐣

For multilane, traffic flow need to be expressed as Where, U represents Speed corresponding to density
vehicle per unit time per lane but due to the lack of K; Uf and Kj represents free flow speed and Jam
lane discipline in most of the developing countries Density respectively. This model satisfies both
having mixed traffic condition, flow is expressed as boundary condition (at k = kj, u=0 and at k = 0, u = uf).
vehicles per unit time per total width of road (Arasan
and Krishnamurthy, 2008) and so does density which Greenberg Model
is also expressed in terms of total width. Greenberg (1959) observed traffic flow in the north
Heterogeneous Traffic and Passenger Car Unit tube of the Lincoln Tunnel at New York City and
developed a logarithmic speed-density model,
The real world consists of vehicles having varying assuming that the traffic flow along a roadway can be
physical as well as operating characteristics rather than considered as a one-dimensional fluid. The
stream of uniform vehicles. Assuming homogeneous mathematical form of Greenshield model is given by:
traffic stream makes calculations much simpler but 𝐾
𝑉 = 𝑉𝑐 × 𝑙𝑛 ( 𝑗)
doesn’t reflect the real scenario. The dividing line 𝐾
between homogeneous and heterogeneous is not clear Where, V represents Speed corresponding to density
cut. Arasan and Krishnamurthy (2008) suggest K, Kj and Vc represents jam density and constant
heterogeneous traffic to exist when the percentage of representing speed corresponding to capacity. This

2|Page
model shows strong goodness of fit under congested maximum flow and free flow speed respectively.
conditions but does not satisfy the boundary condition Drake model generally yields a best fit than other
at low density regime i.e. fails to predict free flow models for uncongested conditions, but fails to
speed. provide a good fit for congested conditions.
Underwood’s Exponential Model Pipes-Munjal model
Underwood (1961) proposed another exponential form Pipes proposed a model with introduction of new
of macroscopic traffic flow model, so as to overcome parameter (n) to provide a more generalized modeling
the drawback of Greenshield model. The Underwood approach shown by the following equation. When n
model is expressed mathematically as: =1, Pipe's model resembles Greenshields' model. Thus
by varying the values of ‘n’, a family of models can be
(−𝐾⁄𝐾 )
𝑉 = 𝑉𝑓 × 𝑒 0 developed.
Where, V represents speed corresponding to density K
U = Uf ∗ {1 − (K )n }
K, Vf and Ko represent free flow speed and optimum j

density (density corresponding to the maximum flow). Where symbol have their usual meanings
This model gives better fit than the Greenshields and
Greenberg models for the uncongested traffic Polynomial Model:
conditions, but fails to do so for congested conditions. Polynomial models of order 2,3,4 and onwards are
It doesn’t satisfy the boundary condition at high calibrated based on the data and are checked for their
density regime. goodness of fit and other estimate. Density and speed
Underwood Model with Taylor Series Expansion relationship expressed as quadratically (polynomial of
order 2) is expressed as:
Exponential function of Underwood model can be
expanded in a Taylor series for the numerical u = uf+ b*k + c*k2
approximation of jam density, which can’t be obtained Where, b and c are additional model parameters, to be
from conventional Underwood model determined.
−k⁄k k k2 k3 k4 Modified Greenberg Model
u = uf e c = uf (1 − + 2− 3+
k c 2k c 6k c 24k 4c
Ardekani and Ghandehari (2008) proposed a Modified
k5
− + ⋯) Greenberg model introducing a non-zero average
120k 5c minimum density, Ko, assuming that there is always
Considering up to third degree of k yields some vehicles on the freeway.
−k⁄ k k2 k3 kj + ko
u = uf e kc = uf (1 − + 2 − 3) u = uc ∗ ln( )
k c 2k c 6k c k + ko

The above equation gives the realistic estimate for the Where Uc and Ko represents speed corresponding to
jam density (kj) by substituting u = 0 maximum flow and Average minimum density
respectively. This model yields a finite free flow speed
Drake Model (Bell-Shaped Curve Model) of Uf=Uc* ln(1 + kj/ko) when density approaches zero.
Drake (1961) improved Underwood model, by Density at capacity of this model will be kc≈ 0.4kj, a
proposing bell shaped model after analyzing all model value close to that of the classical Greenberg model,
from statistical point of view. He estimate density (kc≈ 0.368kj) obtained by solving for dq/dk= 0.
from speed and flow data, fitted the speed vs. density The Drake Model with Taylor Series Expansion
function and transformed the speed vs. density
function to a speed vs. flow function to validate his Conventional Drake model can be expanded using
model. The mathematical expression is: Taylor series expansion to obtain a numerical
approximation for the jam density, as follows:
𝐾2
V = 𝑉𝑓 exp [− 2∗𝐾 2 ] k2 k4 k6
u = uf (1 − 2k2 + 8k4 − 48k6 )
𝑐
c c c
Where, V represents speed corresponding to density k,
Kc and Vf represents density corresponding to
operating characteristics of vehicle moving on
5. Methodology shoulder lead them to be excluded from the volume
count. The PCU value NRS 2070 was adopted.
A. Site Selection
Table 1: Passenger Car Unit (PCU) value as per NRS 2070
The traffic and physical characteristics is almost
similar throughout the Jadibuti – Suryabinayak road Vehicle Bike Car/ Micr Bus Light Truck
section, and after proper inspection, the section just jeep obus truck
ahead of Radhe Radhe chowk, between Sallagahri
intersection and Sankhadhar Shakhwa Chowk was PCU 0.5 1 1.5 3 1.5 3
selected. The considered segment of road was not value
perfectly straight but yet, the curvature was not that Density value was extracted for every 30th second.
steep enough to cause reduction of traffic speed. As all The average value of 15 minute interval was termed as
three data; speed, density and flow were targeted to average density and was expressed as PCU/200m/total
extract from the same video; one kilometer distance width. As the different mode have different range of
can’t be considered. Taking lesser distance might not speed, the result obtained from random sampling may
be realistic and considering larger distance may lead be misleading, and thus stratified random sampling
to parallax error while extracting speed and density based on the volume proportion of different modes
data, thus after trade-off 200m stretch, as separated by was considered. To ensure the representative sample,
two Speed Limit traffic sign was considered as 10% of total volume of each mode observed in 15
reference points. There were two lane, each of 3.5m in minute interval was considered for sampling and
each direction along with 2m shoulder width. finally mean space mean speed (Vs) was calculated.
The vehicle arriving at every 30th second irrespective
of mode and lane was considered for speed calculation
and time taken to pass the considered section was
noted down, until required sample size was achieved.
If the required sample for a particular mode was
completed then that mode is excluded and the other
remaining mode was considered for speed calculation.
If the extracted sample was less than the required
sample then the vehicle arriving at every 15th and 45th
second was considered for data extraction and the
Figure 1: Site selected for study process was continued.
B. Data Collection and Extraction Methodology
Among two directional data, Kotheshwor bound
Video graphic survey, as a part of data collection direction data was used for model calibration and
method, had been conducted for six days, covering Shallagari bound direction data was used in validation
both peak and off peak period including saturday, from process. Twelve hour of data, altogether 48 data (each
mornings 8 am to evening 7 pm. The video camera was representing data corresponding to 15 minute period)
placed at the third floor of the Laxmi Steel building had been extracted for calibration process covering all
and video was recorded and were analyzed to extract range of time. As twelve hours of extracted data covers
the required three basic traffic flow parameter, viz. almost all range of speed, flow and density data,
average speed, flow and density. The road geometry further extraction of data was considered ineffective;
including number of lanes, lane width, shoulder width so 12 hours of data was used for calibration. Six hour
and total width were noted and length were measured data, i.e. 24 points were extracted for validation
via tape. purpose.
Classified traffic volume (categorized into two
wheeler, car, micro, bus, tipper/light truck and truck) 6. Data Analysis
were counted for every 15 minute interval as
suggested by Highway Capacity Manual and was A. Model Calibration
expressed as PCU/15 minute/total width, as suggested
Extracted data was analyzed, based on which various
by Arasan and Krishnamurthy (2008). Insignificant
macroscopic traffic flow models were calibrated.
number of non-motorized vehicles and different
Various conclusions drawn from the models are

4|Page
summarized on Table 2 which gives the estimate of 70
Observed data
free-flow speed (Vf), jam density(Kj), Speed at 60
capacity(Vc), Density at capacity(Kc) and goodness of Underwood curve
50
fit (R2 value).
40
As the section of road is characterized by maximum V = 66.365 e-

Speed (kmph)
30 K/21.516
speed limit of 60kmph and strict enforcement, the free
flow speed will be around 60 kmph or slightly above. 20 R2 = 0.815
This is further supported by the fact the 85th percentile 10
speed of mixed traffic condition was found to be 58
kmph. The length of car is around 3m and they opt to 0
have about 2m clearance. This yield around 40 0 20 40 60 80 100
Density (PCU/200m)
passenger cars in the considered 200m section, but
with 2 lanes, the section considered may accommodate Figure 3: Calibrated Underwood Model
80 PCU/200m in each direction.
60
Drake Model
70 Curve
Observed data 50
60 Observed data
Greenshield curve
40
50
Speed (kmph)

Speed (kmph)
40 30 V = 56.729 Exp (- K2/
30 V = 61.97 - 2.01 * K 305.737)
20
R2 = 0.797
20
10
10
0 0
0 4 8 12 16 20 24 28 32 0 10 20 30 40
Density (PCU/200m) Density (PCU/200m)

Figure 1: Calibrated Greenshields model Figure 4: Calibrated Conventional Drake Model


60
80 Greenberg curve Drake with Taylor…
50 Observed data
70 Observed data
60 40
Speed (Kmph)

Speed (kmph)

50 30 V = 56.71*(1-K2/306.84+K4/188304.85-
40 K6/173339891.3)
V = 14.986 ln (168.61/K) 20
30 R2 = 0.845 10
20
10 0
0 2 4 6 8 10 12 14 16 18 20 22 24
0
Density (PCU/200km)
0 50 100 150 200
Figure 5: Calibrated Drake Model with Taylor
Density (PCU/200km)
Series Expansion
Figure 2: Calibrated Greenberg Model
90 Two degree
polynomial fit
80 Observed Data 70 Underwood with Taylor
70
60 Expansion
Speed (kmph)

60
Observed points
50 50

Speed (kmph)
40 V = 0.4018*K2 - 7.7448*K + 81.696 40
30 R² = 0.8778
20 30 V = 66.16*(1-K/21.8+K2/948.6-
10 20 K3/61976.96
0 R2 value = 0.814
10
0 2 4 6 8 10 12 14
Density (PCU/200m) 0
Figure 6: Calibrated Two degree polynomial model 0 10 20 30 40
Density (PCU/200m)
60 Pipes-Munjal model (N=2)
Observed Data Figure 8: Calibrated Underwood Model with Taylor
50 Series Expansion
40
Speed (Kmph)

30 V = 55.834*[1-
20
(K/19.29)2]

10
0
0 2 4 6 8 10 12 14 16 18 20 22
Density (PCU/200m)
Figure 7: Calibrated Pipes – Munjal Model for N = 2.
Table 2: Calibration of various macroscopic traffic flow model
Model Calibrated Equation Vf Vc Kj Kc R2
(kmph) (kmph) (PCU/200m) (PCU/200m)

Greenshields V = 61.97 (1 – K/30.83) 61.97 30.98 30.83 15.42 0.80

Greenberg V = 14.99 Ln (168.61/K) - 14.99 168.61 62.03 0.84

Conventional
V = 66.36Exp ( -K/21.52) 66.36 24.4 2 - 21.52 0.82
Underwood

Underwood with V = 66.16 * (1- K/21.8+K2/948.6-


66.16 22.05 34.81 21.8 0.82
Taylor Expansion K3/61976.96)

Conventional Drake V = 56.73Exp (- K2 / 305.737) 56.73 34.41 - 12.36 0.75

V = 56.71*(1-
Drake with Taylor
K2/306.84+K4/188304.85 - 56.71 34.26 22.13 12.39 0.75
Series Expansion
K6/173339891.3)

Pipes-Munjal(N=2) U = 55.83 * {1 – (K/19.29)2} 55.83 37.22 19.29 11.14 0.72

Polynomial (N = 2)
V = 0.4018*K2 - 7.7448*K + 81.696 81.69 - - - 0.88

6|Page
Table 3: Re-calibration of recommended models based on validation data

Model Re-Calibrated Equation Vf Vc Kj Kc R2


(kmph) (kmph) (PCU/200m) (PCU/200m)

Greenshields V = 62.93 (1 – K/30.11) 62.93 30.11 31.46 15.06 0.87

Greenberg V = 12.94 Ln (290.33/K) - 12.94 290.33 106.81 0.89

Conventional
V = 64.94 Exp (-K /23.59) 64.94 23.89 - 23.59 0.88
Underwood

Despite high goodness of fit (R2 = 0.88), two degree jam density, assuming 2 kmph as speed at jam density
polynomial model is excluded from the list of best fit it yields jam density of about 75 PCU/200m.
models as it doesn’t yield reasonable estimate beyond
Taking realistic jam density and free flow speed and
the present range of density and also the free flow
reasonable goodness of fit as parameter, Underwood
speed doesn’t seems realistic. Similarly, due to low
model seems to be the best fit model. But new
goodness of fit and unrealistic jam density Pipes-
approach of validation was adopted and all three
Munjal, Conventional Drake and Drake model with
suitable models viz. Greenshield, Greenberg and
Taylor series Expansion doesn’t seems to be the
conventional Underwood were considered for
appropriate models. As both conventional Underwood
validation.
and Underwood with Taylor series expansion yield
same result, conventional B. Validation of Model
Underwood was only considered while validation. As Three recommended models were validated using
illustrated above, all three model except Greenberg Shallaghari bound directional data. Validation process
model (which doesn’t yield any finite free flow speed) was conducted in two phases. First, all three
give more or less a realistic value of free flow speed recommended calibrated models were checked for
(around 60kmph). The jam density in Greenberg their goodness of fit based on validation data. And
model is 168.61 PCU/200m, which seems too high. secondly, recommended models were again
Though Underwood model doesn’t yield any definite recalibrated based on validate
d data and their corresponding goodness of fit, jam terms associated for the recommended models based
density, density at capacity, speed at capacity and free on validation data. Considering goodness of fit,
flow speed were analyzed. After screening through Greenberg model is marginally better than other two,
these two process, finally the best fit model is but the estimates of jam density, speed and density at
recommended. capacity doesn’t seems to be realistic. Greenshield
model provides comparatively better result than the
Table 4: Validation of calibrated models.
Greenberg model for free flow speed and speed at
capacity but the prediction of jam density and density
Model Calibrated Equation R2 St.
Value Error
at capacity is questionable. Now coming to the
conventional Underwood model, all the values i.e. free
𝐾 0.868 0.978 flow speed, speed and density at capacity seem to be
Greenshield 𝑉 = 61.97 (1 − ) realistic and thus it is recommended as the best fit
30.83
model for the study area.
168.61 0.9 0.841
Greenberg 𝑉 = 14.99 × 𝑙𝑛
𝐾
C. Recommended Best Fit Calibrated Model
0.878 0.942 Based on both calibration and validation data,
Underwood 𝑉 = 66.36 × 𝑒 −𝐾⁄21.52
Conventional Underwood model was calibrated and
recommended as the best fit model.
Table 4 reflects that the goodness of fit of all three
recommended models lies within the same range.
Table 3 summarizes the goodness of fit and other
80 Recommended Model fit References
Observed data
Siamak A. Ardekani, Mostafa Ghandehari and Shiva M. Nepal
Speed (Kmph)

60
(2011). Macroscopic Speed - Flow Models for
Characterization of Freeway and Managed Lanes. The
40 University of Texas, Arlington, USA.
V = 66.22 * Exp(-K/21.75)
20 J Roux and C J Bester. Speed- Flow Relationship on Cape Town
R2 Value = 0.83
Freeways. University of Stellenbosch, Department of
Civil Engineering, Private Bag X1, Matieland, 7602.
0
0 10 20 30 40
Density (PCU/200m/7mwidth) 50 Masatu L.M. Chiguma (2007). Analysis of Side Friction
Figure1: Recommended Calibrated Conventional Impacts on Urban Road Links. Doctoral Thesis in Traffic
and Transport Planning, Infrastructure and Planning,
Underwood model Royal Institute of Technology Stockholm, Sweden.
Underwood, R.T. (1961). Speed, Volume, and Density
7. Summary and Conclusion relationships, Quality and Theory of Traffic Flow. Yale
Bureau of highway traffic, page no.141-188.
Van Aerde, M. (1995). Single Regime Speed-Flow-Density
Various conventional macroscopic traffic flow Relationship for Congested and Uncongested Highways.
models; Greenshields, Greenberg, conventional Presented at the 74th TRB Annual Meeting Washington,
Underwood, Underwood model with Taylor series D.C. Paper No. 950802.
expansion, Polynomial model, Pipes-Munjal model, Michel Van Aerde and HeshamRakha. Multivariate
conventional Drake and Drake model with Taylor Calibration of Single Regime Speed-Flow-Density
series expansion were calibrated for this research Relationships. Department of Civil Engineering, Ellis
Hall, Queen’s University, Kingston, Ontario, Canada
work. After screening via calibration and validation K7L 3N6.
process, conventional Underwood model gives far
better and realistic estimate of free flow speed, speed Ardekani S., Ghandehari M. A Modified Greenberg Speed-flow
and density at capacity and approximation of jam Traffic Model. Technical Report 357, Dept. of
Mathematics, University of Texas at Arlington, USA,
density compared to other models. 2008.
Both directional data, which were used in calibration V. ThamizhArasan and G. Dhivya. Simulation of Highly
as well as validation process were combined to Heterogeneous Traffic Flow Characteristics.
calibrate the best fit Conventional Underwood model Transportation Engineering Division, Department of
and yielded a convention Underwood model V = 66.22 Civil Engineering IIT, Madras, Chennai
* Exp(-K/21.75) with 83% strong fit, free flow speed Khan, S.I. and Maini, P., (2000). Modeling Heterogeneous
of 66.22 kmph and density at capacity of 21.75 Traffic Flow. Transportation Research Record, No. 1678,
PCU/200m/7m width. The calibrated model can be page 234- 241.
used for predicting the future scenario of speed and Dr. Hassan Mahdy. A Speed-Flow Model for Austrian
density with anticipated change in traffic volume. Motorways. Public Works Department, Faculty of
Based on developed model, we can have an idea about Engineering Ain Shams University, Cairo, Egypt.
the level of service offered by the facilities. Transport Research Board. (2000). Highway Capacity Manual.
Washington D.C.: Transport Research Board.

8. Further Research Area Government of Nepal. (2070). Nepal Road Standard 2070.
Ministry of Physical Infrastructure and Transprot, Depart of
Roads.
Due to various constraints this research only covers
the calibration of traffic flow model for the freeway
section. Therefore it is recommended to study the
macroscopic traffic flow model for urban roads, within
Kathmandu valley. Similarly lane specific
macroscopic traffic flow model is recommended for
future study. Assessment of present and future level of
service offered based on present and anticipated future
traffic should be conducted with use of the
recommended traffic flow model.

8|Page

You might also like