You are on page 1of 5

2006 IEEE International Conference on

Systems, Man, and Cybernetics


October 8-11, 2006, Taipei, Taiwan

Decision making in transport infrastructures


Zofia Lukszo, Luis Ferreira and Jos Vrancken

Abstract-This paper is about decision making in transport freedom of any decision-maker may be determined by
infrastructures, an area in which most often several actors are others. Next to this, the multi-objective character of
involved, each with its own interests, in which decision decision problems with a variety of criteria, that often may
making tries to optimize different and often conflicting turn out to be conflicting, add an extra dimension to the
criteria simultaneously, and in which decision making has to complexity of the decision process.
take place at different levels, distinguished by geographical
extent and time scale. We distinguish several classes of
When looking at transport infrastructures, decision
solution methods. A number of cases illustrate how the making can take place at many levels in the management
solution methods are being applied in practice. There we see a and operation of infrastructure, and it can range from the
tendency to combine methods and to support decision making capacity utilization of national complex transport
by means of computer-based simulation, where the different infrastructures down to urban traffic management and
levels are reflected in different simulation tools. further down to decisions by individual drivers or
travellers.
I. INTRODUCTION In this paper we will discuss various approaches to
T ransport infrastructures have undergone significant distributed decision making in the transport sector.
1 changes during the past years due to changes in the
market structures, regulations and global II. DISTRIBUTED DECISION SYSTEMS
competition in pricing and quality. The management of the In distributed decision making systems the objectives
operation of these infrastructures is so complex, that to and constraints of any decision-maker may be determined
guarantee the efficiency and effectiveness on the shorter in part by variables controlled by other authorities. In some
and longer term, a well-defined decision-making process is situations, one decision maker may control all variables
an absolute requisite. that permit him to influence the behaviour of other
Let us start with a characterization of infrastructure. The decision-makers as in traditional hierarchical control. In
physical network of an infrastructure and the actors that are essence, each actor is a decision-maker responding to
involved, collectively form an interconnected complex external information: using this information as input he
network where the actors determine the development of the determines his own response. The extent of the interaction
physical network, and the physical network structure may vary depending on the particular environment and
affects the behaviour of the actors [1]. An infrastructure time dimension: in some cases decision makers might be
seen as a socio-technical system represents a multi-level tightly linked, in others they may have little effect on each
system: a hierarchically interconnected or decentralized other.
system, often with local objectives which are related to the For decision making in such systems two important
overall system performance in a rather fuzzy way, optimization methods can be used, i.e. multi-criteria and
especially since the overall objective is often not defined in multi-level optimization [2].
detail and it deals with a long(er) time horizon. The local Multi-criteria (also called multi-objective) problems are
objectives are more detailed and concemed with a short(er) characteristic for the situations when an improvement in
time horizon and/or smaller area of interest. Very often, any one of the criteria involved may be accompanied by a
each hierarchical level in the decomposed system considers worsening in others. Mostly, four classes of solution
a different time horizon, and the activities defined at the methods for multi-objective problems can be distinguished
same system level consider different space boundaries. [3]:
Moreover, in decentralized decision making the degrees of * methods based on some measure of optimality
* interactive methods
This work was supported by the Next Generation Infrastructures * methods searching for Pareto-optimal solutions
Foundation wwwxngjpta.1 and the Delfi Research Centre Next * lexicographic methods.
Generation Infrastructures.
Z. Lukszo (corresponding author) is an associate professor at the Methods based on a measure of optimality make an
Faculty of Technology, Policy and Management of the Delft University of attempt to measure alternatives in one way or another, by
Technology, P.O. Box 5015, 2600 GA Delft, the Netherlands (e-mail: weighing each objective and then optimizing their
zolia atbm.tude1ft.nl). weighted sum or by replacing multi-objective optimization
L. Ferreira is a professor at the School Urban Development of the
Queensland University of Technology, Brisbane, Australia (e-mail: by optimizing only one criterion with the greatest
I ferreira(wgut.edL ai). preference. Therefore, methods of this category translate a
J. Vrancken is an assistant professor at the Faculty of Technology, multi-criteria problem into a single criterion. Such methods
Policy and Management of the Delft University of Technology, the are often supported by computer-based simulation tools.
Netherlands (e-mail: josvtbmJudelttnl).

1-4244-0100-3/06/$20.00 ©2006 IEEE 912


The second group of methods, i.e. interactive methods, model has been adopted in some countries, notably in
uses the information obtained from the decision maker in Great Britain, Germany, the Netherlands and Sweden, with
an iterative process to assign appropriate importance varying degrees of success.
levels, e.g. weights, to all individual objectives. During the Under the vertically integrated model, operators and
solution process the user learns about the nature of the track owners tend to have a customer-service provider
problem under consideration by the comparison, time after relationship. The infrastructure provider exists to service
time, of two sets of multi-objective values corresponding the needs of its client(s). The latter may consist of several
to two different decision vectors as in the method of business units such as passenger services and various types
Woods [4]. Pareto methods of the third group use the of freight services. In some cases, each business group
notion of Pareto optimality to achieve a balance between 'owns' its own track segments, which are divided amongst
objectives. Here the optimal solution appears to be the operators on the basis of major user. User charges may be
natural extension of optimizing a single criterion, in the levied to non-main users using an internal cost transfer
sense that in multi-objective optimization any further system designed to achieve accountability and 'value for
improvement in any one objective requires a worsening of money' outcomes.
at least one other objective. The lexicographic methods It is argued that one of the drawbacks of the vertically
assume that the individual objectives may be ranked by integrated model is its inability to readily and fairly
their importance, so that a sequential optimization of the accommodate new entrants in the form of operating
ordered set of criteria is possible. competitors, sharing a common track infrastructure. The
In a multi-level optimization problem several decision terms and operating conditions of track access need to
makers control their own degrees of freedom, each acting extend to train dispatching rules. This is particularly
in a sequence to optimize their own objective function [5]. important under single line operations, where the train
This problem can be represented as a kind of leader- conflict resolution rules need to be fair and equitable to all
follower game in which two players try to optimize their operators, as well as economically sound [11]. The role of
own utility function F(x,y) and f(x,y), taking into account a certain mediators to ensure 'fair play' is essential in this
set of interdependent constraints. The leader optimizing F case. Government-commissioned agencies are usually
defines an optimal x, so that this term for the objective employed to regulate the operation impartially.
function of the follower is constant, and f(x,y) may be The objectives of the rail operators and owners of
replaced by f(y). However, due to the iterative structure of railway infrastructure may conflict because they can have
this decision process, y can still be represented as a different stakeholders and levels of accountability [12].
function of x. Railway services operated for profit will be concerned with
It should be stressed that even in the linear case with two reducing operating costs and increasing revenues (via
levels the bi-level programming problem is a non-convex growth in market share and/or freight rate increases).
optimization problem which is NP-hard and therefore very Market share increases are closely related to the level of
hard to solve. Moreover, general multi-level programming service that each operator can offer. In this respect, transit
models with an arbitrary number of levels, in which the times and reliability of arrivals are important indicators.
criteria of the leader and the follower can be nonlinear Both these two levels of service attributes are associated
and/or discrete, are most difficult and challenging. with track infrastructure design and maintenance standards.
Therefore, the ability of an operator to perform efficiently,
III. ILLUSTRATIVE EXAMPLES and to gain market share cost-effectively, is closely pegged
This chapter presents four topical challenging examples to its ability to strike an effective contractual arrangement
of distributed decision making in the world of transport with the infrastructure owner.
infrastructures Railway infrastructure owners have to plan and
manage their assets according to their overall strategic
A. Rail Infrastructure Charging objectives. In the case of public ownership of railway
In rail infrastructures, the infrastructure owners and infrastructure, there is an obligation to make investment
managers on the one hand and rail operators on the other decisions which take into account the interests of current
hand need to agree on a set of charges for usage which service operators (sectional/private interest) and the
maximise profit for each main player subject to safety and community to whom the entity is accountable
other operating constraints set by regulators and (collective/public interest). If the infrastructure is to be
governments. The latter may have their own objective owned on a purely commercial basis, the owner has a
function to maximise net community benefit, in terms of profit maximising strategy which will of necessity
transport performance, environmental impact and energy disregard the community costs and benefits of management
consumption. decisions.
Two main railway ownership models have emerged in When trains are scheduled on a rail corridor, the
practice, namely the vertically integrated railway with or objective is to achieve a given level of customer service
without separate internal business units; and the vertically whilst minimising overall operating costs. Customer
separated railway with track infrastructure managed and service in this context is made up of several attributes
owned separately from multiple operators. The separation which include overall journey time and train arrival

913
reliability. In the context of freight movements, the optimize system performance by choosing the optimal tolls
benefits of improved reliability need to be estimated on a for a subset of links, within realistic constraints and subject
train-by-train basis. Each train is usually loaded with to the dynamic route and departure time choice, that is, the
passengers or freight from a range of customers and origin- travel behavioural part.
destination flows.
The elasticity of demand with respect to transit time C. Land use Configuration for Transit Oriented
reliability will differ for each customer, commodity, and Development Projects
origin-destination combination. However, reliability of Transit Oriented Developments (TOD) are higher
arrivals is one of the critical performance measures for all density mixed use residential and commercial
rail markets. The ability of rail systems to compete developments set within walking distance of key transit
effectively, relies to a large extent on this level of service nodes such as rail or bus stations or around activity centres
attributes, as well as on price. such as major shopping centres/offices. TOD aim to
encourage increased ridership in public transport, to
B. Road Pricing as an Instrumentfor Capacity efficiently integrate land use and transport, and to create
Management integrated liveable communities, [13].
Capacity management in a transport sector deals with The main benefits of TOD include:
the allocation of the scarce network capacity of nodes and - Improved mobility and management of travel
links to users. To achieve more balanced capacity demand
utilization in time and space, intelligent road capacity - Higher quality of life: better places to live, work,
allocation principles, such as dynamic road pricing, can and play
support spreading transport demand. In such a system price - Healthier lifestyle: more walking and cycling
levels for tolls can vary over space and time dynamically - Viable economic development: resource efficiency
depending on the traffic conditions in the network and the & effectiveness (e.g. energy, water)
policy objectives of the road authority. A challenging - Reduced pollution and environmental impact.
question is what kind of operational models are needed to
generate optimal distributed dynamic pricing schemes. In Therefore the evaluation of TOD proposals needs to be
this section we consider a network design problem in consistent with a general evaluation framework for other
which the aim is to determine a set of (time-varying) road- public transport projects [14].
pricing levels on road segments on a transportation The development potential TOD sites cannot be fully
network. realised unless there are sound practices which facilitate
The aim of the road authority is to optimize system TOD outcomes, such as: increased public transport usage;
performance by choosing the optimal tolls for a subset of efficient energy and water resource usage; and improved
links, within realistic constraints and subject to the environmental performance [15]. Issues to be addressed
dynamic route and departure time choice, that is, the travel include: appropriate mix of housing, employment, services
behavioural part. More about the problem definition can be and facilities; economic viability of development; energy
found in [2]. In this problem, which can be represented as a and water resource needs; and local urban development
bi-level programming problem, we can distinguish two and demographic contexts. The financial, institutional,
different submodels: the dynamic road pricing model of the legal and regulatory tools and techniques supporting TOD
road authority and dynamic route and departure time need to be put in place.
choice model of travellers, see Figure 1.

Upper level Lower level Local Priva e lanning


'leader' Tolls Iflowers' Community Developer Agency
Objective 0 ectives Objectives

Charge of tavel
behavior
TOD Configurations I
Aim:
aI I
Aim: I
To optimize network To optimize their TOD Outcomes:
performance individual utility improved mobiliand management of travel demand
Figure 1: Schematic representation of the bi-level decision problem in Higher quality of life:better places to live, work, -and play
road pricing problem Healthier lifestyle: more walking, cycling
Viable economic development. resourcetfficiency & effectiveness
The upper level (road authority) describes the overall road Reduced pouton andtenvironmenta impact

performance and the lower level the user-specific objective WINNIMIll Im III 111 ommmimomm

(utility) function. The aim of the road authority is to

914
Figure 2: Iterative TOD planning process more detailed view shows that differences also exist: e.g. a
very low speed limit would virtually eliminate fatal
Decision making tools for engaging multiple actors in accidents but network performance would then aggravate
evaluation and implementation of TOD sites need to lead unacceptably. There is a tendency to solve the multi-
to the optimum development potential for individual TOD criteria aspect by means of a cost/benefit-analysis, where
sites, as well as maximising net environmental, social and each of the criteria is represented monetarily. Still weight
economic outcomes. Figure 2 shows the overall iterative coefficients for the different aspects remain the subject of
planning process that addresses individual actors' preferred continuous debate among the many societal groups and
configurations to reach a preferred compromise solution organisations involved. Therefore shifts in priorities can be
that optimises system wide net benefits for a specific site. observed regularly.
D.Road Traffic Management The most advanced decision support can be found at the
level of RTM-operators in traffic management centres
Road traffic management (RTM) offers a well where in fact all of the aforementioned classes of solution
developed example in complex decision making [6]. In methods are being applied. RTM-operators work at the
RTM decision making takes place at many different levels levels of road segments to regions in the network [8].
(table 1): from the supra-national level (in Europe the Often they are dedicated to a single network (motorway
European Union Traffic and Transport Policy) and the operators, urban operators, rural operators). Still the
national level down to the level of road traffic operators in different networks constitute different spheres of interest
traffic management centres and even down to individual that are often synergic but sometimes opposite: urban
drivers. The road infrastructure is generally considered the operators want to send traffic to the motorway belt road,
most problematic of all societal infrastructures, with whereas motorway operators try to keep the belt road fluid
tremendous costs in terms of human life (in Europe some by reducing the inflow from the urban network by means
40000 fatalities per year) and environmental damage, in of so-called ramp metering. This example also illustrates
addition to the economic costs of congestion. For this that local measures, such as ramp metering, have their
reason a great effort is spent, in research as well as in limitations. The problem can only be dealt with properly at
policy making, to improve this infrastructure, among other the network level, taking the whole region into account.
things by means of improved RTM [7,8,9,10]. This is the reason why there is a strong tendency,
especially in heavily congested areas, to switch to region-
Societal goals oriented RTM [8]. Today, this is made possible by the
Control strategies availability of traffic simulation tools. This kind of tools is
Control tactics used off-line in the preparation of so-called scenarios:
Scenarios series of coordinated local measures for specific, regularly
Link layer measures occurring traffic patterns, thus achieving a network-level
Signals effect by means of local measures. More and more, traffic
Driver behaviour simulation is also used on-line, to evaluate in a very short
Table 1: Levels in RTM time the effect of proposed measures in response to the
sudden occurrence of unforeseen situations or to tune the
The decision making levels are reflected in the road application of the off-line developed scenarios [10].
infrastructure: the road system consists of different Computer based simulation allows for various multi-
networks (motorway, urban, rural) and within each criteria evaluation functions to be applied to proposed
network three different levels can be distinguished: the RTM-approaches. In addition, the RTM-process is
network-level, the road segment level and the point (or essentially interactive, with the operators constantly
cross-section) level. monitoring traffic and dealing with unforeseen situations.
At the national and supra-national level, policies apply What can not be handled by computer simulation, is often
to the whole road system and are intended for many years. handled in a lexicographic way, by prioritizing the
At the other end of the spectrum, many of the decisions by different criteria. Finally, the negotiations between the
individual drivers, such as speed choice or lane choice, are different authorities involved try to find Pareto-optimal
taken at the point level and on the time scale of seconds to solutions to regional traffic problems.
minutes. An interesting development of the past few years, a
At the national and supra-national level, policy making development which takes place largely autonomously, but
is typically multi-criteria decision making, with safety, which is also more and more stimulated by authorities, is
network performance, environmental damage reduction the improvement of the decision making at the lowest
and travel time reliability being the most common criteria. level: the individual drivers. Vehicles are more and more
In addition there are several interdependencies between equipped with computing and communication devices,
road traffic policy and other policy areas, such as space which allow drivers to optimize their trips by means of
planning and economic development. Roughly speaking, real-time traffic data. This kind of user-optimization turns
the factors mentioned are synergic to some degree (e.g. out to be quite effective, even for network-level
network performance improves by fewer accidents) but a optimization. For instance, it causes an automatic

915
spreading of traffic over the network and automatic [14] L. Ferreira, Public transport: economic investment appraisal,
reroutingg in case of accidents or road maintenance. Technical Report, School Urban Development, Queensland
reroutin s .
Authorities are stimulating this by offering
t
increasingly
increa University of Technology, Brisbane, 2005.
[15] Transportation Research Board, Transit oriented development:
better information sources. Developing a strategy to measure success, NCHRP Research
Results Digest 294, Washington, D C, February 2005.
IV. CONCLUSIONS AND FURTHER RESEARCH
Transport infrastructures offer many examples of
settings in which multi-actor, multi-objective and multi-
level decision making is daily practice and applies well-
defined and thoroughly researched methods. Virtually all
of the methods for complex decision making are applicable
in managing transport infrastructures. What we observe in
practice is that different methods are being applied in
combination, that computer-based simulation plays an
increasing role and that computerized systems have
strengthened the role of the lowest and most distributed
level of actors in the hierarchy of decision makers, the
individual consumers.
The relevance of optimization methods and simulation
as decision-supporting tools is very high for many players
in the world of transport infrastructure. For further research
it would be interesting to study these trends in other
infrastructures, such as in the energy and the tele-
communications sectors and to see what lessons the
different infrastructures can learn from each other in the
area of complex decision making.

REFERENCES
[1] Z. Verwater-Lukszo, I. Bouwmans. Intelligent complexity in
networked infrastructures, IEEE conference on Systems, Man and
Cybernetics, Hawaii, October 2005.
[2] Lukszo, Z., D. Joksimovic, M. Bliemer, Optimization of the
operation of infrastructures, IEEE Int. Conf. Networks, Sensing, and
Control, Ft. Lauderdale, Florida, April 2006.
[3] Z. Verwater-Lukszo, A practical approach to recipe improvement
and optimization in the batch processing industry, PhD Thesis,
Eindhoven University of Technology, 1996
[4] D.J. Woods, An interactive approach for solving multi-objective
optimization problems, PhD Thesis, Rice University, 1985
[5] J.F. Bard, Practical Bi-level Optimization, Kluwer Academic
Publisher, 1998.
[6] J.L.M. Vrancken, O.C. Kriise: Intelligent Control in Networks, The
Case of Road Traffic Management, IEEE-ICNSC2006, Ft.
Lauderdale, April 2006.
[7] O.C. Kriise, J.L.M. Vrancken: Architecture for Distributed Traffic
Control, submitted 2006-1-5, ITS World Congress 2006, London.
[8] Rijkswaterstaat: Handbook for Sustainable Traffic Management,
2003, Rotterdam.
[9] R.v. Katwijk, P.v. Koningsbruggen, Coordination of traffic
management instruments using agent technology, Transportation
Research Part C, 2002, pp. 455-471.
[10] MADAM is a traffic simulation model, part of the OmniTRANS
system. Information at: http://www.goudappel.nl/Site/basicsite.nsf/
wwwVwContent/12omnitrans.htm?OpenDocument.
[11] L. Ferreira, Rail track infrastructure ownership: Investment and
operational issues,. Transportation, 24 (2), pp 183-200, 1997.
[12] M. Lake, L. Ferreira, Minimising the conflict between rail
operations and infrastructure maintenance, Transportation and
Traffic Theory in the 21 st Century: Proceedings of the 1 5th
International Symposium on Transportation and Traffic Theory. M
Taylor (Ed.) Elsvier, Oxford, UK, pp63-80, 2002.
[13] Dittmar Hank and Gloria Ohland (ed.), The new transit town : best
practices in transit-oriented development, Washington D.C; Island
Press, 2004.

916

You might also like