The Future of Road Trafﬁc Management: From Local to

Network-wide, from Reactive to Proactive

Jos Vrancken

Delft University of Technology, Delft, The Netherlands

Abstract. This paper describes the current status and future of Road Trafﬁc

Management and the key challenges to achieve this future. It proposes a way

forward that covers both the trafﬁc management theory development, the devel-

opment of supporting systems and the necessary standardization effort needed to

make large scale deployment of Network Management feasible.

1 Introduction

Road trafﬁc is the most important means of transport, both for people and goods, but

also the most problematic one. There are no prospects that this will change funda-

mentally in the coming decades. On the contrary, with increasing standards of living

in many countries in the world, car ownership will continue to grow in an already, at

many places, oversaturated network. The heavier the use of a roads network, the more

reason there is to manage trafﬁc, even though, in dense conditions, it is also hardest

to do this effectively. We can observe each day that just self-organization, although it

certainly plays a important and increasing role, is not enough to prevent congestion.

Trafﬁc management is necessary for an efﬁcient use of the network near its capacity.

The majority of trafﬁc management measures, especially the automated measures such

as trafﬁc signals and ramp metering, are stil local measures with a scope of at most sev-

eral hundred meters. Local measures have obvious limitations: they have a tendency to

solve a congestion problem within their own scope by shifting it to a neighbouring area.

Moreover, without looking at the network context, it is hard to look into the near future,

because the future of one place in a network is, to a large extent, determined by the

current status of trafﬁc elsewhere in the network. Proactive trafﬁc management, trying

to prevent problems rather than solving them, is hard when looking only locally. So,

network management (NM), which is the activity of managing trafﬁc in larger areas,

is urgently needed, even more so as networks become more and more overloaded. But

denser trafﬁc is also trafﬁc in which events at one point have much wider consequences,

which means that a network with denser trafﬁc is a more complex system and harder

to manage. The more NM is needed, the harder it is. In order to understand the true

nature of the problem of developing NM, one must observe that the problem contains

a number of, mutually dependent, components. It’s not only about trafﬁc management,

but also about system development, both control systems and decision support systems

for operators, about the deployment of NM-systems in practice and about measuring

Vrancken J. (2011).

The Future of Road Trafﬁc Management: From Local to Network-wide, from Reactive to Proactive.

In Proceedings of the 1st International Workshop on Future Internet Applications for Trafﬁc Surveillance and Management, pages 93-101

DOI: 10.5220/0004473400930101

 SciTePress

the effectiveness of NM. In the following section we will describe the current status of

NM, other relevant developments in the transport domain, the main reasons why devel-

opment of NM is inevitably a slow process, and ﬁnally we will propose an approach to

move forward in this ﬁeld. The key issues that will be addressed are:

– a framework to describe the problem of control in a network

– the problem of partitioning a network in order to make it manageable

– the problem of assigning priorities to network parts in order to support control

strategies

– the necessary standardization of interoperability among trafﬁc control systems, in

order to make deployment of NM systems practically feasible

2 Current Status of Network Management

Network management is growing slowly but steadily. The subject is best developed in

the area of coordination of trafﬁc lights control systems in urban areas. Research in

this area dates back to the seventies and nowadays, many products are available that

offer various degrees of coordination over more or less extended, urban areas ([10]).

It is well developed, but it is restricted to typical urban trafﬁc and only one kind of

control measure. Integrated Corridor Management ([7]) considers virtually all kinds of

control measures available but it considers only a linear kind of network, a so-called

corridor, the main trafﬁc artery in an area. An example of a system that considers mixed

networks and various types of control measures, together with advanced support for

managing control scenarios, is the SCM system (Scenario Coordination Module, [14],

[15]) in the Amsterdam area. This system was installed in September 2010 and is nowin

the middle of the process of evaluation and full-scale deployment. In [5], an interesting

systems-oriented approach is described that aims at integrating current trafﬁc control

systems into a network control system in Sweden. Thus we can conclude that network

managementdoes take off, but progressis slow. Mature, full-scale network management

systems, covering all networks in an area and coordinating all local measures, are not

yet in operation to the best of our knowledge, let apart that it has been proven that these

systems have a positive effect on network performance.

3 Other Developments in Road Transport

Several other developments in Road Transport are relevant for trafﬁc management. We

mention only a few:

3.1 Cooperative Systems

Cooperative systems are about the increase in data processing and communications on

the road, with connections between vehicles and between vehicles and the roadside.

This is highly relevant to trafﬁc management in several ways. First of all, this develop-

ment increases the process of self-organization among vehicles. Especially at the micro

level, cars can improve trafﬁc properties by means such as adaptive or connected cruise

control. In addition, through the communications facilities, drivers are much better in-

formed about trafﬁc conditions on their intended route. Second, the vehicles become an

important source of trafﬁc data. Suppliers of navigation equipment, such as TomTom in

The Netherlands, are already using this kind of data from their connected devices, for

the purposes of providing their customers with real-time trafﬁc information. To some

extent, they are inﬂuencing trafﬁc in a way that might be called trafﬁc management

(or even network management) but a private party that depends on its customers, will

necessarily manage trafﬁc for the user optimum and not the system optimum that gov-

ernmental trafﬁc management authorities focus on. It is therefore not to be expected

that this ”private” trafﬁc management will eventually replace the trafﬁc management

authorities. Cooperative systems offer interesting capabilities to supply the necessary

sensors and actuators for NM, so instead of making trafﬁc management unnecessary,

they greatly extend the facilities for this activity.

3.2 Cross-sectoral Connections

A second important development might be summarized as the increasing level of con-

nection between the road transport system and other systems in society, both within

transport and in other sectors. Examples of this are:

Multimodal Trafﬁc. This development is in essence about connecting road transport

with the public transport system (both on the road and other forms, such as trains,

streetcars, etc.). This is about ﬁnding optimal trips, in terms of travel time and cost,

by combining different modes of transport (private car, bus, train, shared car, ...);

The Energy Sector. Road transport is highly energy consuming and, by its very na-

ture, strongly dependent on liquid fuels. Much research goes into the introduction

of electric vehicles, and the required charging infrastructure, but convincing break-

throughs are still to come. The extra logistics needed for electric driving, the dif-

ferent driving properties, and the stronger emphasis on driving limited distances,

multimodal trips and car sharing will certainly have an impact on the characteris-

tics of trafﬁc and will change trafﬁc patterns.

Car Sharing. Car sharing is based on the fact that most cars are idle for 95% of the

time, which is of course tremendously inefﬁcient in terms of resources that went

into building so many cars, the cost of ownership and the parking place occupied

by so many vehicles. Car sharing is an obvious solution to this but it requires exten-

sive information processing in order to discover nearby shared cars that are avail-

able for use, and for the administrative handling of trips such that costs are correctly

attributed to the users of a shared car. Car sharing will thoroughly change the re-

lationship between a car and a driver, as the car is no longer his or hers. It is no

longer considered an extension of private space. It looks more like public transport.

This in turn will pave the way towards more outside inﬂuence on driving behavior

and route choice, which offers interesting leads for trafﬁc management. Spreading

trafﬁc over the network, for instance, is a lot easier if one can inﬂuence the route

choice behavior of a fair percentage of vehicles.

4 What is so Hard about Developing Network Management?

However obvious the need for NM may be, its development is slow. The blanket reason

for this is the high complexity of trafﬁc in a network. A single crossing with trafﬁc sig-

nals can already be very complex, let apart a large network containing urban, motorway

and other types of roads. A number of different factors contribute to NM’s complexity,

and the list below is in no way exhaustive.

Chicken-egg Dependencies. There are several chicken-egg dependencies in the prob-

lem of NM development that greatly hamper quick progress in this area. One of

them is the mutual dependency between equipment and network management the-

ory. In order to develop NM theory, we need real-life examples. Such examples re-

quire adequate equipment installed in a given area. The roadside equipment that is

available was not designed for the purposes of NM. But the investment in adequate

equipment is high and should be guided by insight into NM theory. A comparable

dependency can be found in the education of operators. NM will strongly depend

on human operators in trafﬁc management centers, simply because humans have

many capabilities that are very hard or impossible to automate. These operators

have to be trained. But as long as we don’t know for sure how to do NM, it re-

mains hard to train operators. Without trained operators, we don’t get the real-life

examples needed for theory development. It can safely be concluded that trafﬁc

operators play a crucial role in the development of NM, a role that is still often

underestimated.

Measuring Effects of NM is Hard. For those cases in which we do have some degree

of NM installed in real life, it turns out that it is hard to measure its effects on

network performance. Such effects are inevitably statistical effects, averages over

longer periods. They apply to many different trafﬁc states (a speciﬁc approach to

NM is likely to be more effective in one state than in others), which is a second

reason why we need longer periods of measurement, in order to see some differ-

entiation in the data for different trafﬁc states. But networks change during longer

periods, for instance due to road works. And also, trafﬁc demand patterns change

during longer periods. Both are reasons that make the results of one measurement

difﬁcult to compare or to combine with other measurements at different periods in

time. More on this problem can be found in [11].

Legacy Systems. Systems for NM will inevitably consist for a large part of existing,

legacy control systems. As mentioned before, these systems were not designed to

function within an NM system, but in addition to this, it is often hard to connect

with these systems. In practice, most of the effort of installing NM systems is spent

on connecting to legacy systems, as experience with the SCM system mentioned

above has shown. Often these systems are owned by different manufacturers, which

may be competitors of the NM system manufacturer. The willingness to cooperate

is therefore not always obvious.

Sensors and Actuators. However difﬁcult it may be to connect with roadside equip-

ment, in many cases one will have to cope with the absence of such equipment.

The activity of NM is greatly complicated by the fact that it has to be mapped onto

the existing equipment. It is always difﬁcult to install extra equipment for purposes

that are essentially experimental, not only because of the cost involved, but also

because of the necessary permits, the interruption it may cause to trafﬁc, privacy

objections, etc. From the point of view of NM, the existing equipment is usually

scattered more or less randomly over the area under consideration.

Simulation of networks is Hard. Trafﬁc management research heavily relies on mod-

eling and simulating trafﬁc. The models and simulators currently available are not

very scalable. They have difﬁculty to cope with larger areas, already for perfor-

mance reasons. Moreover, models and simulators are often purpose speciﬁc, or at

least have to be conﬁgured in a purpose speciﬁc way. In order to study effects of

network management realistically, calibration data is needed, but the lack of real-

life NM systems makes that such data is virtually non-existent. To make things

worse, trafﬁc itself has a strongly chaotic nature, due to the fact that it is strongly

inﬂuenced by its environment, which contains many chaotic, unpredictable effects,

so much so that the ”regular” situation is a notion in trafﬁc management which is

often used but equally often contested. Predictions based on simulation are most

needed in dense trafﬁc, but as mentioned above, dense trafﬁc in a network happens

to be the most complex, hardest case to manage or to simulate. Dense trafﬁc in a

large area is particularly sensitive to the chaotic inﬂuences from the environment.

This chaotic behavior of trafﬁc applies not only at very short time scales, but also

on larger time scales, in which, for instance, demand patterns change for reasons

outside of the transport domain.

5 The Approach to Developing Network Management

Just like the original chicken-egg problem, the problem of NM development will have

to be solved in a very gradual, evolutionary fashion. Essential steps and corresponding

milestones in the approach are:

Network Partitioning We need a way to partition a network into parts, such that the

control problem for the network as a whole, can abstract from the control problem

inside the parts. For purposes of scalability, this partitioning will have to be done

recursively. In some approaches, a network is split up into a number of main arter-

ies, each artery containing several junctions. Such split-ups are not area-oriented

but artery-oriented. In our opinion, this approach is not a good idea. With such a

split-up, it is obvious that one can do a further split-up of an artery into pieces, such

that each piece contains at most one junction. If two arteries cross each other and

share a junction, this would lead to two systems controlling one network element

(the junction), which would have to be resolved by giving one of the two arteries

priority over the other, which would then effectively be split up. In this way, one

obtains an area-oriented split-up in which control is easier.

Control Framework. We need a control framework to organize the many units of con-

trol present in an NM system, in order to guarantee that control scopes do not over-

lap, and that each control unit is governed by only one parent system. A proposal

for this has been made in [13]. More research is needed on the control of distributed

systems.

Priorities. It is obvious that some parts of a network are more important than others,

in the sense that congestion in important road segments has more impact than in

less important parts (f.i.: a belt road is more important than a residential area). This

plays a crucial role in network management which necessarily must be focused on

keeping ﬂuid important parts at the expense of less important parts. Therefore, an

effective priority assignment to network parts is needed, which reﬂects the impor-

tance that network parts have and which can be used in control strategies.

System Architecture. NM will have to be supported by automated systems. Although

still many details of effective network management are not yet clear, such systems

have to be built, lest there will be no progress in NM theory development. Theory,

systems, and operator experience will have to evolve together. Systems will have

to cooperate, which can be facilitated by a systems architecture which, at the ar-

chitectural level, explains how the trafﬁc management theory can be expressed in

monitoring and control systems and decision support systems for human operators.

Human Operators. Human operators will play a key role. Because of the chaotic na-

ture of trafﬁc, the activity of trafﬁc management is more like stock trading than like

controlling an automated production process with few unpredictable disturbances.

The capabilities of experienced human operators in assessing trafﬁc patterns and

in responding to unpredicted situations with no precedents, are not so easy to au-

tomate. In the course of time, some of these capabilities will be automated, but on

the other hand, the capabilities of operators will also be strengthened by better vi-

sualization of trafﬁc states and better decision support. NM will remain a matter

of humans + machines in the foreseeable future, thus operators should have strong

involvement in the development process of NM.

Interoperability Standard. In order to make real-life deployment of NM systems fea-

sible from the point of view of cost and effort, and also from the point of view of

market reform, an open interoperability standard would be most welcome. Due to

the many connections an NM system needs with both the local control systems it

coordinates and with the neighboring NM-systems, such a standard should support

both horizontal and vertical communication.

6 Key Issues in Network Management

In this section we will go further into a number of the key challenges mentioned in the

approach above.

6.1 Abstraction and Network Partitioning

Abstraction and Network Partitioning both aim at reducing the complexity of NM.

There is an obvious candidate for abstraction, which is also a basis for network par-

titioning. This candidate is the property that, when considering a subnetwork S, only

the boundary, which consists of a integer number of entries and exits of S, is of rele-

vance to the rest of the network. All internal behavior can be ignored, only what happens

on the boundary counts. This abstraction principle is also reﬂected in the so-called net-

work fundamental diagram ([2], [4], [1]) which relates total ﬂow through the boundary

to the number of cars inside the subnetwork. Although mathematically attractive, such

a strong reduction to a single number maybe somewhat too strong. It holds only if some

degree of homogeneity can be assumed for the network, which in a rectangular streets

and avenues network may be a sensible assumption, but for European cities is not very

realistic. There are other, more informativequantities, such as OD-matrices for ﬂow and

for travel time, that are still pure boundary quantities but contain far more information,

and therefore make fewer assumptions about the subnetwork involved. This abstraction

principle implies a way of partitioning a network: namely such that boundaries between

parts are always cross sections (’points’) of carriageways in one driving direction. This

partitioning can be applied recursively, resulting in a tree-structure split-up of the net-

work, similar to (and partially coinciding with) the political partitioning of the surface

of the earth [8] into countries, provinces, districts, etc. The big challenge however is

how exactly to do this for the purpose of effective NM. A number of heuristic rules can

be given, such as the high cohesion, low coupling principle and the rule that bound-

aries should be positioned in quiet areas. In the examples that we know of, the split up

has been done by hand, but this is a tedious and error-prone job, and it makes the split

up not very ﬂexible. Some people suggest that the partitioning should be trafﬁc state

dependent. So for special events or sudden accidents, it should be possible to adapt

the partitioning on the spot. For these reasons, it would be most welcome if effective

partitioning of networks can be automated. There is some research on this subject [9],

[6], [3], [8], which among others points out that many network algorithms tend to be

NP-hard, and that the network structure has a big inﬂuence on its performance, but a

convincing breakthrough in this area is still to be found.

6.2 Priorities

However obvious it may seem that some parts of a network are more important than

others, it is still largely unknown how to assign priorities to network parts, and how to

use these priorities in control strategies. In addition, priorities are also assigned to kinds

of trafﬁc, such as public transport. Both types of priorities will play an indispensable

role in future NM, but still a lot of research is needed on the topic of ﬁnding effective

and implementable algorithms to assign priorities to network parts.

6.3 Interoperability Standardization

An interoperability standard is needed to make deployment of sizeable NM-systems

feasible. Here again, there is a chicken-egg dependency.Such a standard should be able

to express what one control system would like to request from another one. As long

as NM is still not well developed, this remains only very partially known. A recent

initiative for an interoperability standard for trafﬁc control systems, focusing on NM

systems and connections between NM and local control systems, is described in [12].

This approach harnesses the extensibility of XML to solve the chicken-egg dependency

in this challenge.

7 Conclusions

Network Management for road trafﬁc is badly needed in order to cope with our in-

creasingly congested road networks, but its development is slow and its current level of

deployment in practice is still limited. The slow development is caused by the inherent

high complexity of the problem and the chicken-egg dependencies between the various

components of this activity, such as between trafﬁc management theory and the control

systems needed for NM. Important components in the approach that we propose in this

paper are: real-life pilots, algorithms for network partitioning and priority assignment,

architecture and standardization for the systems development needed to support the NM

activity, and strong involvement of operators in the development process.

Acknowledgements. This work was supported in part by the European Community’s

Seventh Framework Programme (FP7/2007-2013)under grant agreementno. INFSOICT-

223844, project C4C, aka Con4Coord.

References

1. Cassidy, M. J., Jang, K., Daganzo, C. F.: Macroscopic Fundamental Diagrams for Free-

way Networks: Theory and Observation. TRB 90th Annual Meeting Compendium of Papers

(2011), Washington, DC 20001 USA.

2. Daganzo, C. F., Geroliminis, N.: An analytical approximation for the macroscopic funda-

mental diagram of urban trafﬁc. Transportation Research Part B: Methodological, 42, (2008)

771–781

3. Etemadnia, H., Abdelghany, K. F.: A network partitioning methodology for distributed traf-

ﬁc management applications. TRB 90th Annual Meeting Compendium of Papers (2011),

Washington, DC 20001 USA.

4. Hoogendoorn, S., Hoogendoorn-Lanser, S;, Kooten, J. van, Polderdijk, S.: Integrated Net-

work Management: Towards an Operational Control Method. TRB 90th Annual Meeting

Compendium of Papers (2011), Washington, DC 20001 USA.

5. Julner, T., Olsson, K.: Deployment of a National Trafﬁc Management System. Proceedings

of the 18th ITS World Congress, 2011, Orlando, Florida, USA.

6. Li, Y., Yang, J., Guo, X., Abbas, M. M.: Urban Trafﬁc Signal Control Network Partitioning

Using Self-Organizing Maps. TRB 90th Annual Meeting Compendium of Papers (2011),

Washington, DC 20001 USA.

7. Marcuson, J. K.: The integration of ICM and ATM. Proceedings of the 18th ITS World

Congress, 2011, Orlando, Florida, USA.

8. Parthasarathi, P., Hochmair, H., Levinson, D. M.: Network Structure and Spatial Separation.

TRB 90th Annual Meeting Compendium of Papers (2011), Washington, DC 20001 USA.

9. Parthasarathi, P., Levinson, D. M.: Network Structure and Metropolitan Mobility. TRB 90th

Annual Meeting Compendium of Papers (2011), Washington, DC 20001 USA.

10. Pooran, F., Martinez, J. C. R.: True Adaptive Control Algorithms A Comparison Of Alter-

natives. Proceedings of the 18th ITS World Congress, 2011, Orlando, Florida, USA.

11. Vrancken, J. L. M., Ottenhof, F., Boel, R. K., Schuppen, J. H. van, Tassiulas, L., Val´e, M.:

WP4 Road Network Control: Design of the Measurement System. Deliverable D-WP4-3 of

the C4C project, http://www.c4c-project.eu/index.php/ﬁledb/showFile/143

100

12. Vrancken, J. L. M., Ottenhof, F., Val´e, M., Lagerweij, R., Goossens, P.: DVM-Exchange,

an interface standard for Trafﬁc Management Systems. Proceedings of the 18th ITS World

Congress, 2011, Orlando, Florida, USA.

13. Vrancken, J. L. M., Schuppen, J. H. van, Soares, M. S., Ottenhof, F.: A Hierarchical Network

Model for Road Trafﬁc Control. Proceedings of the 2009 IEEE International Conference on

Networking, Sensing and Control, Okayama, Japan (2009), 340–344.

14. Yubin Wang, Ottenhof, F., Vrancken, J. L. M.: Integrating Bottom-up Trafﬁc Control into

the Scenario Coordination Module. Proceedings of the ITS Europe Congress, 2011, Lyon,

France.

15. Yubin Wang, Vrancken, J. L. M., Ottenhof, F., Val´e, M.: Next Generation Trafﬁc Control in

the Netherlands. Proceedings of the 18th ITS World Congress, 2011, Orlando, Florida, USA.

101