Behavioural and Network Impacts of Driver Information Systems

Richard Emmerink, Peter Nijkamp

• Publisher: Taylor & Francis
• Year: 2018
• eBook ISBN: 9781351119726
• Edition: 1
• Topic: Economics
• Subtopic: Business General
• Index: Economics

Originally published in 1999, this volume contains a systematic collection of both theoretical and applied studies on user information systems for road users. It is generally expected that reliable information offered to road users will improve the use of scarce capacity on transport networks but from a research perspective the question arises whether the provision of such hard and software will influence the behaviour of road users to such an extent that a more desirable traffic situation will emerge. The book contains European, American and Asian contributions and presents advances and findings in the field of theoretical, simulation and empricial models on driver information systems and behaviour, whilst also paying attention to the design of such systems.

1 Scope of Driver Information Systems

RICHARD EMMERINK AND PETER NIJKAMP

1.1 Prologue

Worldwide, transport is a source of concern. Transportation problems are among the most pressing ones in urban areas. Almost all large metropolitan areas around the world – ranging from, for example, New York, Chicago and Los Angeles in the US to, for example, London, Milan or Paris in Europe, and to, for example, Bangkok, Jakarta and Tokyo in South East Asia – suffer seriously from traffic jams and sometimes even continuously congested roads. Communities and industries located in these areas are strongly affected by these transportation problems. One could think of three different types of problems in these regions: (1) restrictions the congested transportation environment poses upon the economy; (2) social problems related to the limited ability for individuals to move around; (3) environmental impacts in terms of both land use and physical health for the population (see Nijkamp et al. 1998).

The transport sector causes various negative externalities and hence social costs (e.g. traffic jams, environmental decay, fatalities, landscape deterioration etc.). The costs related to the problems caused by congestion have been estimated by various researchers. Although there seems to be some variance in the outcomes, the message conveyed by this research is clear: the costs are very substantial and could be up to almost 0.5 percentage points of GDP, with quite some variance within and between countries (see Verhoef 1996). However, even more alarming than this cost figure is the signal that (1) the congestion costs seem to grow at a rate much faster than GDP, and (2) that traditional solutions to cope with the problem might not be as effective in the future as they were in the past.

In regard to the first point, it is illustrative that congestion costs in the Netherlands have doubled over the past seven years, whereas GDP has only increased with just over twenty per cent in the same period. Concerning the second point, the traditional approach of resolving the congestion problem by simply building more road infrastructure seems to be an unfeasible policy in the years to come for various reasons. Firstly, the costs of expanding the existing road infrastructure might be rather high if all cost aspects are taken into account, i.e. including the social and environmental costs (Boyce 1988; Mogridge 1990). Secondly, in some urban areas with a high population density, it is physically almost impossible to enlarge the current road infrastructure. And thirdly, experience has taught that the congestion relieving impact of building more roads is limited due to latent transportation demand, and the so-called ‘back to the peak’ phenomenon. The latter was, for instance, witnessed in the Amsterdam region after the completion of the ring road (Rijkswaterstaat 1991).

Consequently, research is nowadays largely devoted to finding new effective and efficient policies that might resolve (part of) the congestion problem. Broadly speaking, these new policies may be subdivided into three categories. Firstly, policies that increase the effective capacity of road infrastructure; secondly, policies that impact on the relative attractiveness of various alternative modes of transport; and thirdly, policies that influence driver behaviour and hence the use of the effective capacity of road infrastructure. The first category entails measures ranging from ramp monitoring and the construction of switching lanes, to futuristic measures such as pooling (or bundling) vehicles in order to decrease the space between two subsequent vehicles. The second category encompasses policies such as improving the attractiveness of other modes of transportation that relieve some of the traffic congestion. One could think of improving and/or subsidizing public transport in order to make it more economically attractive for its users. The third category consists of information provision to drivers on traffic conditions and road pricing policies (Ben-Akiva et al. 1991). Both attempt to influence the outcome of the driver’s decision-making process. Due to information provision on the traffic situation or road pricing the driver might be willing to alter his/her route, change his/her departure time, change mode or even postpone the trip. Clearly, the three categories mentioned above are not completely independent. For instance, increasing public transport subsidies, will alter the decisions of some of the road users to use public transportation instead of the own vehicle.

This book focuses on the third category, and more in particular on influencing driver’s behaviour by information provision on the traffic situation.

1.2 Driver Information Systems

The potential of technologically advanced telecommunication and information systems is attracting much attention. Witness, for instance, several projects in the European Union (for example the BATT and MARTA projects within the DRIVE program; DRIVE is an abbreviation for Dedicated Road Infrastructure for Vehicle safety in Europe), the IVHS (Intelligent Vehicle Highway Systems) and ITS (Intelligent Transportation Systems) efforts in the US, the CACS (Comprehensive Automobile Control System) and VERTIS (VEhicle Road and Traffic Intelligence Society) programmes in Japan.

As mentioned above, the main reason for adopting these driver information systems is to reduce travel times and congestion delays. However, it has also been claimed that these systems have a potential to reduce driving stress and anxiety, increase safety and diminish levels of pollution (Mahmassani and Herman 1990; Rumar 1990; Shladover 1993). The contributions in this book, will however be confined to the impact these systems have on congestion, i.e. the main reason for adopting these technologies.

The potential positive effects of information provision to drivers are clear. Driver information is likely to decrease travel times, as drivers are using more information to decide whether, where and when to travel. Its exact impact however, is still rather unclear. This is caused by the potential existence of three adverse effects (1) oversaturation, (2) overreaction and (3) concentration (Ben-Akiva et al. 1991).

Oversaturation occurs if drivers are unable to process the supplied information properly. For instance, information could overload the driver, thus distracting and impairing him from selecting the optimal route. As pointed out by Ben-Akiva et al. (1991), this is mainly a psychological-technical man-machine interaction problem, which can in principle be overcome. Overreaction occurs when drivers’ reactions to information cause congestion to transfer from one road to another. It may also generate oscillations in road usage. Overreaction may happen if too many drivers respond to the same information. Concentration, finally, takes place, if the information provided to the travellers diminishes the natural heterogeneity among the travellers, and makes them more uniform. As a result, a greater number of drivers might select their own best alternative, and consequently drivers with similar preferences will tend to concentrate on the same routes. Thus, more information could potentially generate higher levels of traffic congestion.

1.3 The Models

It is clear that the research issues mentioned in Section 1.2 call for sophisticated analytical tools. There is a variety of technological devices for information systems, there is great diversity among travellers, and there are many site-specific conditions. To study the impact of driver information systems in different situations, the authors in this book will make use of various types of models. Broadly speaking, these can be classified into three categories. Theoretical models, simulation models and empirical models.

1.3.1 Theoretical Models

The theoretical models focus around the equilibrium principle known as Wardrop’s first principle, the user equilibrium. Wardrop (1952) stated that his first principle by ‘the journey times on all the routes actually used are equal, and less than those which would be experienced by a single vehicle on any unused route’ (p. 345). Wardrop’s first principle has been widely accepted as providing a reasonable approximation of the prevailing situation in road transport networks, and it has been used repeatedly in modelling transport demand. In particular, the user equilibrium is widely used in the assignment phase of the traditional four-step modelling process.

There are two equilibrium models widely used in transportation research. First, there is the bottleneck model, originally proposed by the late Nobel laureate Vickrey (Vickrey 1969). The term bottleneck model stems from the fact that Vickrey assumed that traffic congestion is modelled as a queue behind a bottleneck. The main contribution of Vickrey’s bottleneck model is that he endogenized driver’s departure times on the basis of a sound economic principle. He assumed that in equilibrium none of the travellers should be able to decrease travel costs by changing departure time, implying that in equilibrium none of the travellers has an incentive to change. In recent years Arnott et al. (1993) have done much research into the bottleneck model.

Second, the economic model of supply and demand has also widely been used by transport researchers; see for an overview Emmerink (1998). In this model, demand for travel is given by a downward sloping travel demand curve, implying that travel demand will decrease if travel costs increase. Supply for travel is given by curves representing the travel costs for various routes. In equilibrium, travel demand should equal supply, and the travel costs on all routes should be the same, thereby satisfying Wardrop’s first principle.

1.3.2 Simulation Models

The complexity of theoretical models quickly increases as the analysis becomes more realistic. It is then often impossible to have analytic closed-form solutions for the models. In those circumstances, researchers have turned to simulation models. We classify simulation models in two categories. The first category is closely linked to the theoretical models mentioned above. These models simulate theoretical models, where closed-form solutions are not at hand. The second category of simulation models starts from a driver behaviour perspective. These models describe driver behaviour at a micro level, and then simulate what would happen if drivers in a artificial road network would behave in that manner. Examples of the second approach are the simulation models by Mahmassani and associates (see Mahmassani and Herman 1990), and Emmerink et al. (1995).

The disadvantage of simulation models is also its advantage. Complex situations can be modelled, however, analytic solutions are not available. Hence, the danger exists that the model behaves a bit like a black-box. In order to increase the relevance/level of reality of these models, it is important to empirically assess the assumptions on driver behaviour that drive the model. This then brings us to the third category of models.

1.3.3 Empirical Models

As stated above, empirical models are essential in order to render simulation models realistic. In addition, empirical models in itself also can provide evidence pro or contra the effectiveness and efficiency of driver information systems. Among the questions that could be answered by empirical models are ‘how responsive are drivers to driver information?’, ‘which drivers are most responsive to driver information?’, ‘what is the direct measurable impact of certain types of information on congestion levels?’, ‘what proportion of drivers is due to driver information willing to adapt departure time?’, ‘is habitual behaviour affected by driver information?’. Answers to these questions would obviously be very helpful in focusing the research towards the most relevant issues.

Unfortunately, research addressing empirically the behavioural issues of driver information systems has been sparse. This is mainly due to the still very limited implementation of driver information systems around the world, so that rich data is not available. However, as data sources become more available due to the completion of more variable message signs and the larger penetration levels of in-car navigation systems, the future of empirical research in this area seems promising.

1.4 Structure of the Book

Against the background sketched in the previous sections, the present volume aims to present recent advances and findings in the field of theoretical, simulation and empirical models on driver information systems and behaviour, while also next our attention is paid to the design of such systems. Accordingly, the book has four thematic parts which will now concisely be discussed.

The first part (Part A) presents new contributions to the analysis of network impacts of driver information, mainly from a theoretical perspective. This part of the book starts with a chapter by Lindsey on the implications of driver information for the so-called bottleneck model. The author introduces a stochastic version of the Vickrey bottleneck model, in which the number of drivers, N, and the capacity of the bottleneck, s, are random variables that vary unpredictably from day-to-day, but remain fixed during the travel period on a given day. An Advanced Traveller Information System (ATIS) should then provide unbiased information about N and s. Next, the effects of information on two equilibrium regimes are considered: Stochastic User Equilibrium (SUE) and Stochastic System Optimum (SSO). Information is necessarily beneficial for the SSO, but can be welfare-reducing for the SUE. This is because with unpaired congestion, information can induce adverse departure time adjustments. The joint benefits from information and tolling are compared with the benefits from implementing each policy in isolation. Finally, the author introduces route choice and makes also a comparison between the effects of pre-trip information and en-route information.

The second contribution to Part A is offered by Ran and Boyce who address the question of modelling dynamic transportation networks with variational inequalities in an Intelligent Transportation Systems (ITS) environment. This field needs to develop models and algorithms that use real-time data to implement optimal control strategies for traffic, while accommodating both pre-trip planning and en-route travel plan modification. These models must also provide the means for evaluating the benefits of various aspects of ITS. Formulation of an optimization model for a dynamic transportation network problem involves many strict assumptions about travel time functions and constraints. It is usually very difficult to find appropriate objective functions for such formulations. To overcome this general problem, the variational inequality approach is proposed as a general way of formulating dynamic transportation network problems. This chapter aims to summarize a new generation of dynamic network equilibrium models, incorporating dynamic travel choice problems including motorists’ departure/arrival time choice and route choice. These models are expected to be able to function as on-line dynamic travel forecasting and evaluation tools, and eventually as real-time on-line models of urban transportation networks.

The final chapter in the part on theoretical models is written by Kobayashi and Tatano. These authors address the issue of welfare measurement in the context of information and rational expectation in modelling driver information systems. Their contribution presents an analytical framework to measure the economic values of information systems for route navigation. The welfare measurement is made by use of the random expected utility model with rational expectations. Expected consumer’s surplus can be used to evaluate the economic benefits of information systems for route navigation. The authors use various numerical examples to illustrate how these indices can be applied to route navigation problems.

The next part of the book (Part B) is devoted to simulation models for driver information systems. The first chapter on this theme is written by Noland who deals with information in a two-route network with both recurrent and non-recurrent congestion. In his article a two-route network is simulated to determine the behavioural impacts of non-recurrent incidents and information on both route and departure time choice. A demand side model is specified (based on a stated preference survey) that accounts for travel time and scheduling costs. This is interacted within the two-route network to determine the impact of changes in both road capacity and incident probability (i.e. non-recurrent events). This allows the calculation of expected travel time costs for different users of the system and is extended to analyse the impacts of providing users with information on actual travel times. Ex post costs to commuters of obtaining or not obtaining information are analysed as well.

Next, Watling presents a stochastic process model of day-to-day traffic assignment and information. In contrast to conventional static equilibrium assignment models, the author introduces stochasticity as a result of the impact of information provision on driver’s behaviour (e.g. route choice, diversions, etc.). He illustrates next the potential of this model on the basis of a two-link network.

Finally, an analysis of route guidance systems in the presence of responsive signal control policies is given by Clegg and Smith. In their contribution two strategies for finding drivers’ routes through a network are considered. The first is based upon drivers with knowledge of previous days of travel on the network and the second is based upon drivers with knowledge only of conditions on the network at an instant. It is shown that, in conditions of constant demand, drivers who base their route-choice on previous days of travel can experience a lower travel time than drivers who base their route choice on network conditions and take no account of the possible future evolution of traffic on the network. It is further shown that there may be good reasons to believe that route-guidance systems based upon instantaneous knowledge interact favourably with responsive signal control policies to reduce travel time.

It goes without saying that there is a great need for empirically tested, applicable models in the area of driver information systems. Part C of the present volume contains several interesting model applications. The first one is given by Van Berkum and Van der Mede who offer a discussion on driver information and the (de)formation of habit in actual route choice conditions. Their contribution aims to provide a comprehensive discussion of behaviourally relevant aspects of dynamic route choice behaviour in a transport modelling context. The contribution deals with the following questions: What are the relevant features of route choice behaviour to be considered for transport modelling purposes? Which behavioural assumptions must be made to model route choice behaviour in information environments? What is the role of the (de)formation of habit in the choice process in different information environments? Their article forms the reflective introduction to the other papers in this part.

The next contribution is by Bonsall and Palmer, who critically assess potential sources of data on driver r...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• Figures and tables
• Contributors
• Preface
• 1 Scope of Driver Information Systems
• Part A Theoretical Models on Network Impacts of Driver Information
• 2 Effects of Driver Information in the Bottleneck Model
• 3 Modelling Dynamic Transportation Networks with Variational Inequalities
• 4 Information and Rational Expectations in Modelling Driver Information Systems: A Welfare Measurement
• Part B Simulation Models and Driver Information Systems
• 5 Information in a Two-Route Network with Recurrent and Non-Recurrent Congestion
• 6 A Stochastic Process Model of Day-to-Day Traffic Assignment and Information
• 7 Dynamic Simulation of a Simple Route Guidance System in the Presence of Responsive Signal Control Policies
• Part C Empirical Models of Behavioural Change
• 8 Driver Information and the (De)formation of Habit in Route Choice
• 9 Route Choice in Response to Variable Message Signs: Factors Affecting Compliance
• 10 Experimental Analysis of Effects of Travel Time Information on Dynamic Route Choice Behaviour
• 11 Impacts of Pre-trip and En-route Information on Commuters’ Travel Decisions: Summary of Laboratory and Survey-based Experiments from California
• 12 Designing ATIS for Familiar Drivers: Preliminary Behavioural Concepts
• 13 Analysis of Drivers’ Response to Information using Fuzzy Logic and Approximate Reasoning Models
• 14 The Effect of Advanced Traveller Information Systems (ATIS) on Travellers’ Behaviour
• Part D Design Aspects of Driver Information Systems
• 15 The Close Connection between Dynamic Traffic Management and Driver Information Systems
• 16 The Roles of Driver Information and Congestion Pricing Systems
• 17 Probe-based Surveillance for Travel Time Information in ITS