Stochastic Transport in Complex Systems: From Molecules to Vehicles

Stochastic Transport in Complex Systems

From Molecules to Vehicles

First Edition

Andreas Schadschneider

Debashish Chowdhury

Katsuhiro Nishinari

ELSEVIER

AMSTERDAM • BOSTON • HEIDELBERG • LONDON • NEW YORK • OXFORD

PARIS • SAN DIEGO • SAN FRANCISCO • SINGAPORE • SYDNEY • TOKYO

Table of Contents

Cover image

Title page

Copyright page

Dedication

Preface

HOW TO USE THIS BOOK

ROUTE 1: INTRODUCTION TO STOCHASTIC SYSTEMS

ROUTE 2: INTRODUCTION TO THE MODELING OF TRAFFIC PHENOMENA

Acknowledgments

Part One: Methods and Concepts

Chapter One: Introduction to Nonequilibrium Systems and Transport Phenomena

Publisher Summary

1.1 INTRODUCTION

1.2 CLASSIFICATION OF NONEQUILIBRIUM PHENOMENA

1.3 HIERARCHY OF DESCRIPTION AT DIFFERENT LEVELS

1.4 INDIVIDUAL-BASED MODELS

1.5 POPULATION-BASED MODELS

1.6 FLUID FLOW: THEORETICAL DESCRIPTIONS AT DIFFERENT LEVELS

1.7 BACK TO DISCRETE MODELS: MIMICKING HYDRODYNAMICS WITH FICTITIOUS PARTICLES

1.8 PHASE TRANSITIONS, CRITICAL DYNAMICS, AND KINETICS OF PHASE ORDERING

Chapter Two: Methods for the Description of Stochastic Models

Publisher Summary

2.1 QUANTUM FORMALISM

2.2 MEAN-FIELD AND CLUSTER METHODS

2.3 BETHE ANSATZ

2.4 MATRIX-PRODUCT ANSATZ

2.5 OTHER METHODS

2.6 NUMERICAL METHODS

2.7 APPENDICES

Chapter Three: Particle-Hopping Models of Transport Far from Equilibrium

Publisher Summary

3.1 ELEMENTS OF RANDOM WALK THEORY

3.2 ASYMMETRIC SIMPLE EXCLUSION PROCESS

3.3 ZERO-RANGE PROCESS AND EXACT RESULTS

3.4 EXTENSIONS AND GENERALIZATIONS

3.5 PHYSICS OF THE ZRP

3.6 PARTICLE-HOPPING MODELS WITH FACTORIZED STATIONARY STATES

3.7 GENERALIZED MASS TRANSPORT MODELS

3.8 APPENDIX

Chapter Four: Asymmetric Simple Exclusion Process – Exact Results

Publisher Summary

4.1 ASEP WITH PERIODIC BOUNDARY CONDITIONS

4.2 ASEP WITH OPEN BOUNDARY CONDITIONS

4.3 PARTIALLY ASYMMETRIC EXCLUSION PROCESS

4.4 EXTENSION OF THE ASEP TO OTHER UPDATE TYPES

4.5 BOUNDARY-INDUCED PHASE TRANSITIONS

4.6 EXTENSIONS OF ASEP

4.7 MULTISPECIES MODELS

4.8 OTHER RELATED MODELS

4.9 APPENDICES

Part Two: Applications

Chapter Five: Modeling of Traffic and Transport Processes

Publisher Summary

5.1 INTRODUCTION

5.2 CLASSIFICATION OF MODELS

Chapter Six: Vehicular Traffic I: Empirical Facts

Publisher Summary

6.1 MEASUREMENT TECHNIQUES AND DETECTORS

6.2 OBSERVABLES AND DATA ANALYSIS

6.3 FORMATION AND CHARACTERIZATION OF TRAFFIC JAMS

6.4 FUNDAMENTAL DIAGRAM

6.5 METASTABILITY AND HYSTERESIS

6.6 PHASES OF TRAFFIC FLOW

6.7 RAMPS, INTERSECTIONS, AND OTHER INHOMOGENEITIES

6.8 HEADWAY DISTRIBUTIONS

6.9 OPTIMAL-VELOCITY FUNCTION

6.10 CORRELATION FUNCTIONS

6.11 PSYCHOLOGICAL EFFECTS

Chapter Seven: Vehicular Traffic II: The Nagel–Schreckenberg Model

Publisher Summary

7.1 DEFINITION OF THE MODEL

7.2 FUNDAMENTAL DIAGRAM AND LIMITING CASES OF THE NaSch MODEL

7.3 ANALYTICAL THEORIES FOR NaSch MODEL WITH Vmax > 1

7.4 SPATIO-TEMPORAL ORGANIZATION OF VEHICLES

7.5 APPENDICES

Chapter Eight: Vehicular Traffic III: Other CA Models

Publisher Summary

8.1 SLOW-TO-START RULES, METASTABILITY, AND HYSTERESIS

8.2 CRUISE-CONTROL LIMIT

8.3 CA MODELS OF SYNCHRONIZED TRAFFIC

8.4 OTHER CA MODELS

8.5 CA FROM ULTRADISCRETE METHOD

8.6 CA MODELS OF MULTILANE TRAFFIC

8.7 EFFECTS OF QUENCHED DISORDER

8.8 BUS-ROUTE MODEL

8.9 ACCIDENTS

Chapter Nine: Vehicular Traffic IV: Non-CA Approaches

Publisher Summary

9.1 FLUID-DYNAMICAL THEORIES

9.2 SECOND-ORDERFLUID DYNAMICAL THEORIES

9.3 GAS-KINETIC MODELS

9.4 CAR-FOLLOWING MODELS

9.5 COUPLED-MAP MODELS

9.6 OTHER APPROACHES

Chapter Ten: Transport on Networks

Publisher Summary

10.1 NETWORKS AND TRANSPORT

10.2 BML MODEL OF CITY TRAFFIC

10.3 CHOWDHURY-SCHADSCHNEIDER MODEL

10.3.1 Crossroads with Signals

10.3.2 ChSch Model

10.3.3 Traffic Signal Optimization

10.4 HIGHWAY AND CITY NETWORKS

10.5 COMPUTER NETWORKS AND INTERNET TRAFFIC

Chapter Eleven: Pedestrian Dynamics

Publisher Summary

11.1 INTRODUCTION

11.2 EMPIRICAL OBSERVATIONS AND COLLECTIVE PHENOMENA

11.3 CELLULAR AUTOMATA MODELS

11.4 FLOOR FIELD CA

11.5 OTHER MODELS

Chapter Twelve: Traffic Phenomena in Biology

Publisher Summary

12.1 INTRODUCTION

12.2 TASEP FOR HARD RODS: MINIMAL MODEL OF TRANSCRIPTION AND TRANSLATION

12.3 TASEP FOR PARTICLES WITH LANGMUIR KINETICS: MINIMAL MODEL OF KINESIN TRAFFIC

12.4 TRAFFIC IN SOCIAL INSECT COLONIES: ANT-TRAILS

Guide to the Literature

Bibliography

Index

Copyright

Elsevier

Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands

The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK

Copyright © 2011, Elsevier BV. All rights reserved

No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher.

Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone: (+44) 1865 843830, fax: (+44) 1865 853333, E-mail: permissions@elsevier.com. You may also complete your request online via the Elsevier homepage (http://elsevier.com), by selecting Support & Contact then Copyright and Permission and then Obtaining Permissions.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

ISBN: 978-0-444-52853-7

For information on all Elsevier publications visit our Web site at www.elsevierdirect.com

Printed and bound in Great Britain

10 11 9 8 7 6 5 4 3 2 1

Dedication

Dedicated to

Martina, Indrani, and Yumiko

Preface

Historically, statistical mechanics was developed as the molecular kinetic theory of matter. Its formalism for systems in thermodynamic equilibrium has been established on a solid foundation. Twentieth century has witnessed many triumphs of this formalism in solving many mysteries of bulk matter in thermodynamic equilibrium. Encouraged by the overwhelming success of the concepts and techniques of statistical mechanics in understanding the physical properties of inanimate matter, some statistical physicists have taken bold steps to explore virgin territories beyond the traditional boundaries of physics using the same tool box. Some of these systems were subjects of earlier investigation in other traditional disciplines. Such unconventional applications of statistical mechanics have opened up new horizons of interdisciplinary research; one of the most successful among such joint ventures is the area of soft matter.

But, despite relentless efforts of statistical physicists over the last century, an equally strong foundation of nonequilibrium statistical mechanics remains elusive. One class of nonequilibrium systems that has received lot of attention in recent years consists of mutually interacting particles, which are driven by an external field. Such intrinsically nonequilibrium systems can settle down to a nonequilibrium steady state which, in principle, can be far from the equilibrium state that it could attain in the absence of the external field. Many physical systems have been successfully studied in the last two decades, using the techniques developed for models of such field-driven interacting particles.

In more recent times, nonequilibrium statistical mechanics has found even more unusual applications in research on traffic flow of various types of objects. A list of such systems would include not only vehicular traffic, but also traffic-like collective phenomena in living systems, as well as in colonies of organisms. The common modeling strategy is to represent the motile elements (i.e., vehicles or an organism) by self-propelled particles, which convert chemical energy (derived from fuel or food) into mechanical energy required for the forward movement in a dissipative environment. In such generic models, the mutual influences of the motile elements on the movement of each other are captured by postulating appropriate interparticle interactions.

Just as nonequilibrium statistical mechanics has helped in getting deep insight into the general physical principles governing the spatio-temporal organization and traffic flow in a wide variety of complex systems, which were earlier subjects of investigation in other branches of science and engineering, these disciplines have also enriched physics by proving wider testing grounds for the tool box of nonequilibrium statistical mechanics. The problems that we are studying here are classic examples of complex systems. Their collective behavior is not just a simple superposition of the individual behaviors, but the interactions among the particles/vehicles/agents… lead to new features of the dynamics with, sometimes, surprising consequences.

In Part I of this book, we give a pedagogical introduction to the conceptual framework and technical tools of nonequilibrium statistical mechanics. Although our main focus is on driven systems of interacting particles, we present the subject from a much broader perspective. In Part II, we illustrate the use of the formalisms developed in Part I by applying these to vehicular and pedestrian traffic, as well as to traffic-like collective phenomena in biology.

Modern life is crucially dependent on vehicular traffic. Our travel to our work places, to supermarkets, to hospitals, to schools and universities all depend on vehicular traffic. The supply of our daily needs also mostly come by trucks. We are all too familiar with the irritating traffic jams. But, traffic jam is not a modern phenomenon; it was familiar to inhabitants of cities even before the invention of motorized vehicles. Even in ancient cities like Rome, carts and chariots were often stuck in traffic jam and elaborate plans for traffic control were made even by Leonardo da Vinci.

However, study of traffic as a branch of science and engineering started growing only in the second half of the twentieth century. Interestingly, new approaches of investigations in traffic science have been opened up mostly by physicists, perhaps, because of the close analogy between vehicular traffic and systems of interacting particles driven far from equilibrium. The notable among the early contributors include, for example, Montroll, Potts, Prigogine, and others. The conceptual framework and technical tool box of statistical physics has turned out to be extremely useful in modeling traffic phenomena and analyzing the properties of the models.

Nature has dealt with traffic-like phenomena for billions of years. In the recent years, it has become clear that the environment inside a cell is, in many respects, similar to urban traffic system where molecular motors carry cargo over long distances by moving along filamentary tracks. Molecular traffic jam can lead to diseases. Therefore, fundamental understanding of these traffic phenomena will not only help in diagnosis but also in the control of those diseases that arise from malfunctioning of the molecular motor traffic in living systems.

Traffic-like phenomena occur in biological systems at almost all levels of organization- from individual molecules and molecular-self assemblies to cells and organisms. The collective movements of ants toward the food source and their return to the nest appears very similar to vehicular traffic. Surprisingly, quantitative studies of this traffic phenomenon started only a couple of years ago despite the possible applications of the results in various ant-based routing algorithms in communication networks, in swarm intelligence, and even in decentralized management.

In the past, capturing large number of empirical facts and phenomena with a few mathematical equations have always led to great progress in scientific theories. Maxwell unified all the empirical phenomena in electricity and magnetism in terms of the four equations named after him. Similarly, our attempt has been to highlight the conceptual unity among the apparent diversity of systems covering a long range of length scales and time scales. A common conceptual thread runs through all these traffic-like phenomena. We hope this monograph will stimulate further exchange of ideas across disciplinary boundaries of physics, chemistry, biology, as well as technology and engineering enriching all.

HOW TO USE THIS BOOK

In Part I, we present a systematic pedagogical treatment of the theoretical formalisms which we, then, use in the Part II to develop and analyze models of transport and traffic phenomena. This book is self-contained in the sense that all the formal theoretical methods required in Part II are available in Part I.

The formalisms discussed in Part I are essentially based on the concepts and techniques of nonequilibrium statistical mechanics. For the convenience of nonexperts, we minimize subtle technical details in the main chapters of Part I. Instead, for the benefit of readers interested in such details, lengthy mathematical calculations have been given in the appendices.

In Chapter 1, while familiarizing the reader with the inventory of the toolbox of nonequilibrium statistical mechanics, we also mention their potential use in the subsequent chapters in the context of traffic science. In Chapter 2, we train the beginner in using some of the most powerful tools. In particular, we introduce several analytical, numerical, as well as phenomenological approaches for studying (quasi-) one-dimensional driven diffusive systems. In Chapter 3, we focus more specifically on a technique that, in recent years, has been used successfully in the mathematical treatment of many particle-hopping models. In Chapter 4, we give a comprehensive overview of our current understanding of the asymmetric simple exclusion process (ASEP), which is the most important particle-hopping model relevant for theoretical modeling of traffic. In fact, most of the particle-hopping models of vehicular traffic, which are described in Part II, are extensions of ASEP appropriate for capturing some specific traffic phenomena.

We classify the traffic models in Chapter 5 so that the reader does not lose sight of the forest for the trees. We review the well-known empirical facts about traffic phenomena in Chapter 6. In Chapter 7, we discuss all the theoretical treatments of the Nagel—Schreckenberg model, the minimal and most basic model of vehicular traffic on highways, in detail. All the other CA models of vehicular traffic, many of which are related to the Nagel—Schreckenberg model, are considered in Chapter 8. For the sake of completeness of the overview of theoretical approaches to vehicular traffic, we also present the non-CA models in Chapter 9. The models of highway traffic have been modified and extended appropriately to capture some of the key features of vehicular traffic on networks of roads and highways in cities and greater urban areas. We present these models in Chapter 10 along with similar models of internet traffic. Theories of pedestrian traffic and traffic-like collective phenomena in biological systems are now making rapid progress. We provide a glimpse of this fast emerging frontier area in Chapters 11 and 12.

There are at least two different routes how to use this book.

ROUTE 1: INTRODUCTION TO STOCHASTIC SYSTEMS

Part I of this book can be used as a textbook for a specialized course on stochastic transport in systems driven far from equilibrium. In such a course, depending on the academic background and research interest of the students, a few topics from Part II can be selected to show interesting applications of the methodology.

ROUTE 2: INTRODUCTION TO THE MODELING OF TRAFFIC PHENOMENA

Part II of this book can serve as a comprehensive introduction to the models of traffic and traffic-like collective phenomena. A reader, who is mainly interested in applied research in traffic, may need to refer to Part I if and only if (s) he wants to delve into any particular formalism.

A long list of references is provided as a guide to the literature. Therefore, this compendium will also serve as a valuable reference for experts actively engaged in research on traffic models and related phenomena in complex systems.

Acknowledgments

We have benefitted from our interactions with large number of students and colleagues over the last decade. First of all, we thank Ludger Santen who started this endeavor with us. But due to other commitments, he graciously resigned from this project. Several parts of this book are based on our joint review published almost 10 years ago in Physics Reports. We would like to thank all our collaborators with whom we have carried out our original work over the last two decades in this area of research. We thank Robert Barlovic, Aakash Basu, Maik Boltes, Jordan G. Brankov, Elmar Brockfeld, Carsten Burstedde, Debanjan Chowdhury, Mohcine Chraibi, Rashmi Desai, Georg Diedrich, Christian Eilhardt, Bernd Eisenblätter, Nils Eissfeldt, Minoru Fukui, Ashok Garai, Andrej Gendiar, Kingshuk Ghosh, Sascha Grabolus, Philip Greulich, Vishwesha Guttal, Torsten Huisinga, Nobuyasu Ito, Rui Jiang, Alexander John, Jens Kähler, Andreas Kemper, Janos Kertész, Macoto Kikuchi, Ansgar Kirchner, Kai O. Klauck, Hubert Klüpfel, Wolfgang Knospe, Tobias Kretz, Dima Ktitarev, Ambarish Kunwar, Maria Elena Larraga, Jack Liddle, Wolfgang Mackens, Arnab Majumdar, Kai Nagel, Akihiro Nakayama, Alireza Namazi, Lutz Neubert, Tomotoshi Nishino, Stefan Nowak, Yasushi Okada, Abhay Pasupathy, T.V. Ramakrishnan, Vladislav Popkov, Andrea Portz, Andreas Pottmeier, Vyatcheslav B. Priezzhev, Nikolaus Rajewsky, Antonio del Rio, Christian Rogsch, Tobias Rupprecht, Ludger Santen, Michael Schreckenberg, Gunter M. Schütz, Armin Seyfried, Ajeet Sharma, Shishir Sinha, Bernhard Steffen, Robin Stinchcombe, Yuki Sugiyama, Shin-ichi Tadaki, Christian Thiemann, Akiyasu Tomoeda, Tripti Tripathi, Satoshi Yukawa, Peter Wagner, Jian-Sheng Wang, Andreas Winkens, Robinson Will, Dietrich Wolf, Marko Wölki, Frank Zielen, and J. Zittartz for enjoyable collaborations. We apologize to everyone whose name might have been left out unintentionally.

The collaboration of the three authors has been supported, at various stages, by the Deutsche Forschungsgemeinschaft (DFG, Germany), Federal Ministry of Education and Research (BMBF, Germany), Alexander von Humboldt Foundation (Germany), Council of Scientific and Industrial Research (CSIR, India).

Over the last two decades, our ideas on traffic and related phenomena have been strongly influenced by our interactions with our collaborators, as well as with Cecile Appert-Rolland, Mustansir Barma, Richard Blythe, Martin Burd, Bikas Chakrabarti, Bernard Derrida, Deepak Dhar, Audrey Dussutour, Martin Evans, Erwin Frey, Raghavendra Gadagkar, Stephan Grill, Stefan Grosskinsky, Tom Hanney, Dirk Helbing, Joe Howard, Frank Jülicher, Boris Kerner, Stefan Klumpp, Anatoly Kolomeisky, Stefan Krauß, Joachim Krug, Reinhard Lipowsky, Reinhard Mahnke, Satya Majumdar, Gautam Menon, David Mukamel, and Dietrich Stauffer.

However, we, the authors, are solely responsible for any conceptual error that might have crept in.

We are indebted to many students and friends for their constructive criticism of our style of presentation. Special thanks go to Richard Blythe, Mohcine Chraibi, Christian Eilhardt, Philip Greulich, Tomohiro Sasamoto, Gunter Schütz, Armin Seyfried, and Daichi Yanagisawa who devoted lot of their valuable time to read earlier versions of the manuscript and for their comments and suggestions. However, we are responsible for the error(s), if any, which could not be detected before sending the manuscript to the printer.

We could not have completed this book without the full cooperation and support from our family members who had to sacrifice many holidays so that we could devote the time to meet the deadline. Last, but not the least, our most sincere thanks to Kristi Green and Anita Koch at Elsevier for their patience and for extending the deadline many times in the last 4 years.

Part One

Methods and Concepts

Chapter One

Introduction to Nonequilibrium Systems and Transport Phenomena

Publisher Summary

This chapter provides a pedagogical introduction to wide varieties of formalisms of nonequilibrium statistical mechanics. Each formalism is introduced from a broad general perspective. Formalism can be adapted to treat vehicular traffic and other traffic-like collective phenomena. The techniques of modeling from a broader perspective within the framework of the formalisms of nonequilibrium statistical mechanics and nonlinear dynamics of interacting particles and fluids are described. The theoretical models of traffic and traffic-like collective dynamics are formulated using several different mathematical formalisms. The choice of the formalism depends on the level of description. Descriptions at different levels are certainly not equivalent to each other. In principle, the dynamical equations at a higher level can be derived from one at a lower level. Moreover, even at a given level, alternative but equivalent formulations of the dynamics are possible. The traffic-like systems involve purely classical transport phenomena, and quantum mechanics plays no role in these processes. However, there are subtle differences between vehicular traffic and the collective transport of interacting classical particles. Furthermore, the mechanisms of transport by particles of inanimate matter and those driving transport in the living world also differ in details in spite of all the apparent similarities.

Contents

1.1. Introduction   3

1.2. Classification of Nonequilibrium Phenomena   4

1.3. Hierarchy of Description at Different Levels   6

1.4. Individual-Based Models   7

1.4.1. Newton’s Equations, Hamilton’s Equations, and Individual Trajectories   7

1.4.2. Langevin Equation   9

1.5. Population-Based Models   10

1.5.1. Master Equation and Irreversibility   10

1.5.2. Fokker–Planck Equation   13

1.6. Fluid Flow: Theoretical Descriptions at Different Levels   14

1.6.1. Liouville Equation and Flow in Phase Space   15

1.6.2. From Liouville Equation to Boltzmann Equation: BBGKY Hierarchy   16

1.6.3. From Kinetic Theory to Navier–Stokes Equation   17

1.7. Back to Discrete Models: Mimicking Hydrodynamics with Fictitious Particles   19

1.7.1. Driven-Diffusive Lattice Gas Models   20

1.8. Phase Transitions, Critical Dynamics, and Kinetics of Phase Ordering   22

1.8.1. Critical Dynamics: Role of Symmetry and Conservation Laws   22

1.8.2. Kinetics of Phase Ordering: Metastable and Unstable Initial States   23

1.8.3. Phase Transitions in Driven Systems   24

In this introductory chapter we provide a pedagogical introduction to the basic concepts and formalisms of non-equilibrium statistical mechanics and nonlinear dynamics of interacting particles and fluids. Special emphasis is given to those aspects which are relevant for the modelling of traffic and transport systems. Various levels of description and the relation between them are discussed. Furthermore basic collective phenomena and related topics like phase transitions are considered from a rather general perspective.

1.1 INTRODUCTION

This book is mainly concerned with traffic and traffic–like collective phenomena. We shall first present an overview of the techniques of modeling from a broader perspective within the framework of the formalisms of nonequilibrium statistical mechanics and nonlinear dynamics of interacting particles and fluids.

The traffic-like systems that we will consider here involve purely classical transport phenomena, and quantum mechanics plays no role in these processes. However, there are subtle differences between vehicular traffic and the collective transport of interacting classical particles. Furthermore, the mechanisms of transport by particles of inanimate matter and those driving transport in the living world also differ in details in spite of all the apparent similarities [219]. In this chapter, we provide a pedagogical introduction to wide varieties of formalisms of nonequilibrium statistical mechanics [62, 1530]. After introducing each formalism from a broad general perspective, we indicate how the formalism can be adapted to treat vehicular traffic and other traffic-like collective phenomena.

1.2 CLASSIFICATION OF NONEQUILIBRIUM PHENOMENA

is the probability distribution of the configurations. A characteristic feature of the thermodynamic equilibrium of a system at the absolute temperature T is given by the Gibbsian form

   (1.1)

and kB is the Boltzmann constant.

Broadly speaking, we can classify nonequilibrium systems into two groups: (1) systems at or near equilibrium, and (2) systems far from equilibrium (see Fig. 1.1). For systems near thermodynamic equilibrium, Onsager’s regression hypothesis [186] states that the relaxation of macroscopic nonequilibrium disturbances follows the same laws, which govern the regression of spontaneous microscopic fluctuations in a system that is in equilibrium. For such systems, linear response theory holds, and the fluctuation–dissipation theorem connects response and dissipation with correlations in the fluctuations.

Figure 1.1 Equilibrium and nonequilibrium systems in statistical mechanics.

, in general, depends on time t. However, given enough time for spontaneous time evolution, such systems may eventually reach a stationary state (also called steady stateis not necessarily identical to the Gibbsian formula (1.1). It is worth emphasizing that not all stationary states are thermodynamic equilibrium states; the state of thermodynamic equilibrium is a very special stationary state, as we will show later in this chapter.

In fact, for a large class of truly nonequilibrium systems, no Hamiltonian can be defined; such models are defined in terms of the rates of the transitions between different configurations. Such a system may attain a nonequilibrium stationary state or exhibit nonstationary behavior for ever. Therefore, systems far from equilibrium can be divided further into two subclasses: (1) systems for which Hamiltonian can be defined and the stable thermodynamic equilibrium exists, so that the system evolves spontaneously toward this equilibrium; and (2) systems for which neither the Hamiltonian nor the Gibbsian equilibrium state exists. Systems of type (1) can attain thermodynamic equilibrium state after sufficiently long time. But, a system of type (2) may attain only a nonequilibrium steady–state (see Fig. 1.1).

Consider, e.g., a window glass which is a supercooled liquid and is in a metastable state. It can, in principle, attain the corresponding crystalline equilibrium structure after sufficiently long time. But, in contrast, a conducting wire carrying an electrical current can, at best, attain a nonequilibrium steady state after the transient currents die down. Systems in the latter category will be referred to as driven nonequilibrium system where, loosely speaking, an external drive maintains the corresponding current. A difference of chemical potential between the two ends of a sample can drive a current of material particles. Similarly, an electric potential gradient can drive an electrical current.

How do systems evolve from initial states that are far from equilibrium? If an equilibrium exists, such a system would tend to equilibrate. However, such systems follow different mechanisms depending on whether the initial state is metastable or unstable [475]. From a metastable initial state, the equilibration of the system occurs through nucleation of droplets (nuclei) of the stable phase and their growth, as well as coalescence [712]. A system in an unstable state evolves in time by the coarsening of a random interconnected structure leading to the phase separation of the components (spinodal decomposition or Ostwald ripening). We will discuss these issues in more detail in Section 1.8.2.

In a thermodynamic system, i.e., a system for which Gibbsian equilibrium exists, the difference between the metastable and unstable states can be explained easily by considering an appropriate thermodynamic potential with more than one extremum. The maxima of the potential correspond to the unstable states, whereas all the minima, except the lowest one, are identified as metastable. The global minimum corresponds to the stable equilibrium state. The concepts of unstable and metastable steady states of a driven nonequilibrium system cannot be distinguished in a similar manner because analogs of the thermodynamic potentials, which can be defined for all driven systems, are not known. Nevertheless, fluctuation–driven spontaneous transitions from metastable steady states to stable steady states have been observed, for example, in vehicular traffic.

1.3 HIERARCHY OF DESCRIPTION AT DIFFERENT LEVELS

The theoretical approaches of nonequilibrium statistical mechanics can be broadly divided into two categories: (1) individual-based and (2) population-based (Fig. 1.2). The individual-based models describe the dynamics of the individual elements explicitly. Just as microscopic models of matter are formulated in terms of molecular constituents, the individual-based models of transport are also developed in terms of the constituent elements. Therefore, the individual-based models are often referred to as microscopic models. When watched closely, traffic flow appears very similar to particles that are not only driven forward but also influence each other (although direct collisions take place only when they accidentally crash against each other). By exploiting these similarities with systems of interacting driven particles, we can treat traffic problems as those of nonequilibrium statistical mechanics.

Figure 1.2 Individual-based models versus population-based models.

However, if vehicular traffic is watched from a long distance (say, from a hill top or a helicopter), traffic flow at peak hours resembles fluid flow where we do not see each individual vehicle; the fluid is characterized by a local density at different locations along the highway. In the population-based models individual elements do not appear explicitly and, instead, one considers only the population densities (i.e., number of individual elements per unit area or per unit volume). The spatio-temporal organization of the elements are emergent collective properties that are determined by the responses of the individuals to their local environments and the local interactions among the individual elements. Therefore, in order to gain a deep understanding of the collective phenomena, it is essential to investigate the linkages between these two levels of organization.

Usually, but not necessarily, space and time are treated as continua in the population-based models and partial differential equations (PDEs) or integro-differential equations are written down for the time-dependent local collective densities of the elements. The individual-based models have been formulated following both continuum and discrete approaches. In the continuum formulation of the Lagrangian models, differential equations describe the individual trajectories of the elements.

The theoretical models of traffic and traffic-like collective dynamics have been formulated using several different mathematical formalisms. The choice of the formalism depends on the level of description. Descriptions at different levels are certainly not equivalent to each other; however, in principle, the dynamical equations at a higher level can be derived from one at a lower level. Moreover, even at a given level, alternative but equivalent formulations of the dynamics are possible.

1.4 INDIVIDUAL-BASED MODELS

1.4.1 Newton’s Equations, Hamilton’s Equations, and Individual Trajectories

The natural starting point of Newtonian dynamics is to write down the equations of motion for all the individual particles. Suppose, mi is the mass of the i-th particle. Then, for a system of N particles, the Newton’s equations for the individual particles are given by

(1.2)

is the external force on the ileads to coupling of the N differential .

The state of the entire Nat any arbitrary instant of time t. Thus, each such state corresponds to a single point in the 2Nd-dimensional phase space of the N particles in a d-dimensional space. The 2Nd coordinates of the phase space are spanned by Nd positions of the N particles and the corresponding Nd momenta. Therefore, the time evolution of a single N-particle system is described by a trajectory in this 2Nd-dimensional phase space.

Hamilton’s canonical equations of motion for the generalized coordinate q and generalized momentum p of the i-th particle are given by

   (1.3)

is a function of q and p. These equations provide an alternative formulation of the dynamics of the N-particle system that is completely equivalent to the description in terms of the Newton’s equations.

In principle, depending on the available computational resources, the Newton’s equations can be solved numerically for a finite system consisting of N particles provided N is not very large. In those situations where each particle corresponds to a molecule, this method is usually referred to as molecular dynamics (MD). However, precise initial conditions for all the N particles are difficult, if not impossible, to extract from any independent measurements. Furthermore, even a slight error in the initial conditions, which can never be ruled out because of the finite resolution of any measurement, can have enormous effects on the accuracy of the results of the simulations. This is a consequence of the well-known phenomenon that the separation between two trajectories, which initially differ by an insignificant amount, can grow exponentially with time if the equations of motion are nonlinear.

Newtonian Approach to Modeling Traffic

A real piece of matter consists of molecules. Therefore, atomic or molecular constituents of matter can be represented by a particle. What would be the counterpart of a particle in models of vehicular traffic? Although a vehicle consist of an enormously large number of molecules, we can still treat each of these as a particle for appropriate traffic phenomena, depending on the length scales involved in the problem. We should not forget that, while developing his laws of mechanics, Newton treated sun and the planets as point particles which was appropriate for the celestial phenomena he studied with that formalism. Therefore, from now onwards, we will use the more general term particle to denote a motile object (a molecular motor or a living cell or a cell or an organism or a vehicle).

What would be the counterpart of force in the analog of Newton’s equation that models vehicular movement in traffic? In order to get an intuitive clue, note that the Newton equation can be regarded as a linear response relation, i.e., a linear relation between a cause and the corresponding effect. In this case, force acting on a particle is the cause and the corresponding acceleration is the effect, 1/m being the analog of a response coefficient, i.e., a measure of the extent of response to unit force. This is analogous to the linear response relation for ohmic conductors where electric current is caused by an electric field (potential gradient); electrical conductivity being the corresponding linear response coefficient. Thus,

   (1.4)

Based on this philosophy, many years ago, statistical physicists formulated the so-called car-following models of vehicular traffic which will be described in detail in Section 9.4.

The interaction between different vehicles cannot be captured by interparticle physical interactions alone because the drivers respond to the surrounding stimuli. Therefore, for modeling traffic or pedestrian dynamics with Newton-like equations, one sometimes includes social forces in the stimulus while acceleration of the vehicle or pedestrian corresponds to the response. Similar approaches have been used also in formulating the equations of motion of organisms (e.g., ants) for modeling their collective movements. It is worth pointing out here that Newton’s third law does not hold for social forces.

1.4.2 Langevin Equation

The true microscopic description in terms of Newton’s equations (or, equivalently, Hamiltonian’s equations) cannot account for the irreversible phenomena in the macroscopic world because these equations are invariant under time reversal. For a more concrete discussion, let us consider a steel ball falling under gravity in a viscous medium. As we learn in elementary Newtonian mechanics, the particle attains a terminal (steady) velocity because of the viscous drag force. But, if instead of focussing our attention only on the steel ball, we also wrote down the Newton’s equations for all the N particles of the surrounding medium, the system of N + 1 equations would have perfect time reversal symmetry. In other words, the dissipative viscous drag appears only when we study the dynamics in a subspace of the entire phase space of the system by projecting out the other degrees of freedom.

Now, instead of the steel ball, consider a pollen grain which is small by macroscopic standards but is still several orders of magnitude larger than the molecules of water. If the pollen grain is immersed in water, it executes an apparently erratic movement, which is known as Brownian motion. Guided by deep intuitive insight, Paul Langevin wrote down an equation of motion for the Brownian particles which was later named after him [837]. The Langevin equation for a Brownian particle of mass m can be written as

   (1.5)

where the second term on the right-hand side captures the dissipative forces while the third term on the right-hand side is a random force. Since, by definition, the random force is unpredictable, only its statistical properties can be postulated. The simplest thing one can assume is that the average of this random force vanishes so that the average motion is still similar to that of the steel ball we discussed earlier. In his original work, Langevin assumed the form

   (1.6)

for the dissipative viscous force, where γ is a phenomenological coefficient.

Fr(t) is the random force (noise) whose time dependence is not known. However, in order that the average velocity satisfies the Newton equation for a particle in a viscous medium, we further assume for the average

   (1.7)

and the autocorrelation function

   (1.8)

where ξ = Fr/m and, at this level of description, Γ is a phenomenological parameter. The prefactor 2 on the right-hand side of Eqn. (1.8) has been chosen for convenience.

At this level of description, one effectively separates the entire composite system into two subsystems one of which is the motile element, i.e., the system of interest; the other subsystem, which includes the rest of the composite system, is treated as a reservoir or bath. The time evolution of an individual motile element is now given by a differential equation that, unlike Newton’s equation, is both stochastic and irreversible.

Langevin Approach to Modeling Traffic

The formalism we have described so far in this section describes passive Brownian particles in external fields. In the last decade, this formalism has been extended to model active Brownian particles [1278] where one assumes that the dissipative force is such that it is positive at small velocities (and, hence, accelerates rather than decelerating the particle), whereas it is negative for all velocities beyond a critical value.

In the context of vehicular traffic phenomena, the Langevin equation has not been used widely (for an exception, see [is time-dependent. In such cases, one needs an additional equation which governs the slow time evolution of the landscape itself. For example, as we will show later, in the case of molecular motors, the landscape changes periodically and completes one cycle with the consumption of each molecule of the chemical fuel. In case of collective traffic-like movements of ants on trails, the dynamics of the landscape and the forward movement of the ants are coupled in a very interesting manner; we will consider this in detail in Section 12.4.

1.5 POPULATION-BASED MODELS

1.5.1 Master Equation and Irreversibility

at time t . In the context of traffic we will often face situations where time can still be considered continuous, but the configurations of the system form a discrete set. The dynamics of the system corresponds to transitions between the configurations. The master equation is given by [1409]

(1.9)

is the transition probability,. The first and the second terms on the right-hand side are called the gain and loss terms, respectively.

The summations over the configurations have to be replaced by integrations when the configurations form a continuum; for a continuous random variable, z, the master equation has the form

(1.10)

In the commonly used forms (1.9) and (1.10), time t is treated as a continuous variable. In many applications, particularly in modeling traffic, time is discretized in short intervals Δt. In such situations, given the configuration of the system at time t, the master equation yields the corresponding configuration at time t + Δt:

(1.11)

, this can be written in the more compact form

   (1.12)

during the timestep Δtdue to probability conservation. The form of the discrete master at time t + Δ t at time t.

Note that if

   (1.13)

is independent of time (see Eqn. 1.9), i.e., the system is in a stationary state (also called steady statewhich, in turn, imposes the condition

   (1.14)

; this is the condition of detailed balance. Situations may arise in systems where the strong condition (1.14) of detailed balance does not hold, but a steady-state exists because of the pairwise-balance condition

   (1.15)

exists so that the condition (1.15) is satisfied. The differences between the conditions (1.13), (1.14), and (1.15) are explained graphically in Fig. 1.3. Consider the 10 configurations labelled by the integers 1, …, 10. If the only transitions into and out of the configuration 1 are those shown in Fig. 1.3a, then dP1/dt. In other words, for any arbitrary configuration i, stationarity of Pi, i.e., the condition dPi/dt = 0, is satisfied if merely the total probability current into the configuration i is exactly balanced by the total probability current out of i. In contrast, the equilibrium, which is a special stationary state, requires detailed balance (see Fig. 1.3b): the net forward and reverse probability current in between the members of each pair of configurations must balance exactly. Clearly, this is a much stronger condition than what was needed to guarantee a nonequilibrium stationary state. Finally, for many models, a pairwise-balance is also satisfied by the nonequilibrium stationary state (see Fig. 1.3c): The probability current from configuration 1 to 2 is exactly balanced by that from 2 to 3 so that dP2/dt = 0. Similarly, the probability current 2 to 3 is equal to that from 3 to 1, and so on. Such cyclic processes are important for many phenomena, particularly in enzymatic catalysis, which occur in molecular motor transport.

Figure 1.3 Illustration of detailed and pairwise balance.

For a more general discussion of the master equation and its relation with mathematical graph theory, we refer to the classic review article by Schnakenberg [1252]. This has been extended, e.g., to classify nonequilibrium steady states in terms of their stationary probability distribution and the associated currents [1525, 1526] and to derive fluctuation theorems [31, 498].

In a more formal way, the master equation (1.9) can be recast as

   (1.16)

i.e.,

   (1.17)

is given by

   (1.18)

.

We will come back to this formulation in the next chapter where it is the starting point of the so-called quantum formalism.

Master Equation for Traffic Models

The master equation approach has been used extensively for formulating models of vehicular traffic. While time is discretized in most of these models, master equation in continuous time has also been used. This will be discussed in detail in Part II of this book.

1.5.2 Fokker–Planck Equation

The Fokker–Planck equation is a valid description at the same level at which the Langevin description holds. The Fokker–Planck equation can be derived from the master equation under an appropriate truncation of an expansion known as Kramers–Moyal expansion [1409]. The truncated equation is essentially a master equation

   (1.19)

is a second-order differential operator given by

   (1.20)

with the jump moments

   (1.21)

The Fokker–Planck equation has also been used for the description of traffic, see e.g., [899].

1.6 FLUID FLOW: THEORETICAL DESCRIPTIONS AT DIFFERENT LEVELS

Fluid flow has been studied at several different levels of description (Fig. 1.4). It is worth pointing out that not all the intermediate steps of the hierarchy are necessarily essential in a comprehensive treatment. For example, the hydrodynamic counterparts of many lattice gas models can be derived directly without going through the intermediate steps of deriving any lattice Liouville or Lattice Boltzmann version.

Figure 1.4 Relation between the different levels of description.

Therefore, in this section, we provide a brief overview of these theoretical approaches at different levels. A real fluid consists of interacting molecules. Therefore, in principle, one may naively attempt formulating the mathematical theory of a fluid in terms of the Newton equations of its constituent particles. Even if any computer, in forseable future, could handle such enormous number of coupled ordinary differential equations, it is practically impossible to get the states of all the particles at a given instant of time, which can be treated as the initial state of the system. Moreover, even if it were possible, integration of these equations (or, equivalently, MD simulation) would yield enormous amount of information most of which is redundant for describing fluid flow at the macroscopic level.

The large number of molecules in the fluid is actually not a bane but a boon for an alternative approach based on the principles of statistics. Any statistical prediction improves with increase in the sample size! In the following subsection, we outline the general formal approach for deriving the equations for such a statistical description of a fluid.

1.6.1 Liouville Equation and Flow in Phase Space

Newton’s equation describes the time evolution of each individual particle. A given initial condition leads to a unique trajectory in the phase space. But, since the initial conditions are usually not available, let us assume a probability distribution for the individual conditions. Thus, instead of a one single point, now we have a swarm of points corresponding to several possible initial states of the system. This swarm flows in the phase space, each point of the swarm tracing its own trajectory. This alternative, but equivalent, statistical approach is based on the N, which is the joint , and so on up to the molecule N , all at the same time t. The dynamical equation governing the time evolution of the distribution fN is given by

(1.22)

and

(1.23)

The equation (1.22) is known as the Liouville equation. Note that this equation can be recast as

(1.24)

is a 2Nd-dimensional vector in phase space and its components are the positions and momenta of the N particles. The form (1.24) of the Liouville equation reveals its close formal similarity with the equation of continuity for incompressible fluids, the density of fluid corresponds to the probability density fN. Using Hamilton’s canonical equations, the equation (1.22) can also be written as

   (1.25)

is given by

   (1.26)

Newtonian dynamics is invariant under time reversal. The Liouville equation also does not account for irreversible processes. Both contain equally large amount of information about the system lot of which is, for most practical purposes, redundant. Besides, to our knowledge, no Liouville-like equation has ever been written down for traffic flow. Nevertheless, the Liouville equation is a good starting point for systematically reducing the amount of irrelevant information and for introducing elements that account for the irreversibility observed in the macroscopic world. Moreover, it also allows a transparent link between the reversible dynamics, which it describes, and irreversible processes which are described by the master equation.

The form (1.16) of the master equation appears formally similar to the form (1.25) of the Liouville equation. But, there are crucial fundamental differences between the two equations; the Liouville equation captures the reversible Newtonian (or, equivalently, Hamiltonian) dynamics, whereas the master equation describes irreversible processes. To our knowledge, the relation is best elucidated with the help of the Kac ring model [320, 321].

1.6.2 From Liouville Equation to Boltzmann Equation: BBGKY Hierarchy

The Liouville equation provides a kinetic description of a many-body system. It contains as much information as contained collectively in all the Newton’s equations of the N particles. Nevertheless, the Liouville equation forms a convenient starting point for a formal way of systematically eliminating irrelevant information and cast the theory in terms of reduced distribution functions

(1.27)

. The chain of time evolution equations for fM with decreasing M is known as the BBGKY hierarchy (named after Bogoliubov–Born–Green–Kirkwood–Yvon [which can be formally written as

(1.28)

at time t. The right-hand side symbolically represents the effects of binary collisions; this is usually a complicated integral whose actual form depends on the detailed nature of the intermolecular interactions. This is known as the Boltzmann equation; it provides a kinetic description of one body in a many-body system. Derivation of this equation is based on the assumption of molecular chaos (what Boltzmann called Stosszahlansatz) according to which

(1.29)

i.e., absence of correlation between molecules encountering a binary collision. This assumption, however, has profound implications as it breaks the time-reversal symmetry inherent in the Newton’s equations and, equivalently, in the Liouville equation.

Boltzmann-Like Equation in Kinetic Theory of Vehicular Traffic

Extending the basic concepts underlying the Boltzmann equation, Prigogine and coworkers developed a kinetic theory of vehicular traffic. However, several subtle questions had to be addressed in formulating the strategy. First, for a dilute gas, an arbitrary initial distribution relaxes to the corresponding equilibrium distribution (Maxwell’s distribution) after a sufficiently long time. What would be the analog of equilibrium distribution when the same method is applied to vehicular traffic? Second, what would be the counterpart of the Stosszahlansatz for the kinetic theory of vehicular traffic? These questions will be addressed in Section 9.3 while we present the kinetic theory of vehicular traffic.

1.6.3 From Kinetic Theory to Navier–Stokes Equation

Let us first define the averaged variables, namely local mass density, momentum density, and energy density by the equations

   (1.30)

   (1.31)

   (1.32)

where m is the mass of a molecule. In the hydrodynamic regime (where the deviations from thermodynamic equilibrium is small), one gets the standard equation of fluid dynamics at the macroscopic level (for simplicity, we drop the equation (1.32) from our consideration and this is valid for isothermal fluids). The Navier–Stokes equation is given by

(1.33)

where v is the viscosity, P must be satisfied.

The Navier–Stokes equation is a force-balance equation for elementary volumes of fluid. Although small by macroscopic standard, these elementary volumes contain a sufficiently large number of molecules to make the density-based description a sensible approach. Thus, at the level of Navier–Stokes equation, the dynamics of individual constituent particles do not appear explicitly.

Because of the conservation of the fluid, in addition to the Navier–Stokes equation, the fluid density also satisfies the equation of continuity

   (1.34)

is the current density.

The Burgers equation

   (1.35)

is a nonlinear diffusion equation in one-dimensional space. Although the setting up of this equation can be motivated by hydrodynamics, the equation itself is too simple to capture several realistic features of real fluids. For example, it neither accounts for shear nor for vortices in a fluid and its solutions correspond to an infinitely compressible medium!

Fluid-Dynamical Approach to Vehicular Traffic

When viewed from a long distance, the flow of vehicles in a crowded traffic on a highway resembles fluid flow. Naturally, mathematical methods that have been successfully used for fluids have been also tried for the study of traffic flow.

Because of the conservation of the vehicles, an equation of continuity can be easily written down for vehicular traffic. Pioneering work in this direction was done by Lighthill and Whitham [874, 875] more than half a century ago. However, establishing an appropriate analog of the Navier–Stokes equation for traffic flow has been much more difficult. Since vehicles in a traffic are subjected to social forces, in addition to real physical forces, the force balance becomes more subtle than in the case of a real fluid. We will study the fluid-dynamical models of vehicular traffic in Section 9.1.

Some of the most successful particle-hopping models of vehicular traffic reduce to the Burgers equation [87, 167] in the continuum limit. One of the key features of this equation is that two solutions of it can coexist in such a way that u(x) varies from a high value to a low value over a narrow region whose width vanishes in the limit v → 0. Such a shock wave solution, in the limit of vanishing viscosity in the Burgers equation, has a physical interpretation as the front of a traffic jam.

1.7 BACK TO DISCRETE MODELS: MIMICKING HYDRODYNAMICS WITH FICTITIOUS PARTICLES

The Navier–Stokes equation is a PDE for a fluid in the continuum space-time. Over the last few decades, a few alternative approaches have been developed in terms of fictitous particles. The rules of dynamical evolution of these fictitious particles capture only the most essential features of the corresponding real particles. These dynamical rules are prescribed in such a way that an appropriate coarse-graining of the model would yield the corresponding correct PDEs for the density variables that characterize the real macroscopic system in the continuum space-time limit.

Cellular Automata Hydrodynamics

Numerical solution of Navier–Stokes equation (or similar nontrivial equations of hydrodynamics) requires discretization of both space and time. Therefore, the alternative discrete formulations may be used from the beginning. Besides, the theory can be formulated in terms of fictituous particles each of which can, in principle, capture many degrees of freedom of the real fluid, i.e., each particle can be viewed as an object obtained after coarse-graining over many real particles around a point in the fluid.

In these models, fictituous particles are constrained to move along the bonds of a regular lattice following well-defined prescriptions or update rules. These rules mimic both collisions of real particles, as well as their free streaming in between collisions. The update rules govern the evolution of the discrete positions and discrete velocities of the fictituous particles in time which is also discretized.

Lattice Boltzmann Equations and Hydrodynamics

The lattice Boltzmann equation (LBE) is based on the discretization of the phase space to allow a numerical solution of the Boltzmann equation. It may be regarded as an extension of the cellular automata (CA) hydrodynamics, where the dynamical variable associated with each discrete space-time point is a real variable [1317].

1.7.1 Driven-Diffusive Lattice Gas Models

is discrete.

However, there are more than one ways of labeling the configurations. To illustrate this point, let us consider a lattice gas in one dimension. Suppose, the lattice sites are labelled by the integers j such that j = 1 and j = L correspond to the leftmost and the rightmost sites, respectively, of the finite system. Let the dynamical variable σj denote the instantaneous status of occupation of the j-th site, i.e., whether or not site j is occupied and, if so, what is the discrete speed of the particle. Using σj the configuration of the system can be unambiguously specified by the set {σj}. This convention of labeling the states of the system is used in what is commonly called the site-oriented approach or Euler representation. The alternative particle-oriented approach, which is sometimes called Lagrange representation, is closer to the Newtonian description where the state of the system is specified by listing the positions of the particles in the lattice and the corresponding discrete velocities.

All the particle-hopping models of vehicular traffic are essentially lattice gas models in which each particle represents a vehicle. In other words, the size of each cell is chosen to be such that it can accommodate just one vehicle at a time. There are some practical advantages in modeling traffic systems with CA-type rules. It is quite realistic to think in terms of the way each individual motile element responds to its local environment and the series of actions they perform. The lack of detailed knowledge of these behavioral responses is compensated by the rules of CA. Usually, it is much easier to devise a reasonable set of logic-based rules, instead of cooking up some effective force for dynamical equations, to capture the behavior of the elements. Moreover, because of the high speed of simulations of CA, a wide range of possibilities can be explored which would be impossible with more traditional methods based on differential equations.

The forward motion of the vehicles is captured by imposing a drive, which biases the motion of the vehicle. In some models, the bias is introduced by a driving external field, whereas in others, the particles are self-driven. The difference between these two types of models are explained the following subsections.

Field-Driven Particles

To our knowledge, the Katz–Lebowitz–Spohn (KLS) model [715, 716, 1523] is the simplest among the models for systems of interacting field-driven particles. Suppose the variable ni describes the state of occupation of the site i (i = 1, 2, …, N) on a discrete lattice; ni is allowed to take one of the only two values, namely, ni = 1 if the site i is occupied by a particle and ni = 0 if it is empty (or, equivalently, occupied by a hole). The Hamiltonian for the system, in the absence of any external driving field, is given by

   (1.36)

where the summation on the right-hand side is to be carried out over all the nearest-neighbor pairs and J takes into account the corresponding interparticle interactions.

The instantaneous state (configuration) of the system at time t is completely described by ({n}; t). For example, in case of a system of a one-dimension system of length L one has ({n}; t) ≡ (n1, n2, …, nL; t). Similarly, for the Lx × Ly . The average density ρ of the particles is given by

   (1.37)

where Ns is the total number of available sites, e.g., Ns = L for a linear chain or Ns = LxLy for a square lattice of size Lx × Ly. Note that, because of the conservation of the particles, the density ρ is conserved by the dynamics.

The dynamics of the system is governed by the well-known Kawasaki dynamics: at any nonzero temperature Twhere β = (kBT)−1 (kB is the difference in the energy of the new and old configurations while ℓ.

For the KLS model with attractive interparticle interactions (J > 0) on a square lattice, there is not only an ordered state at all T < Tc(E), but also the critical temperature Tc(E) increases with E, saturating at a value Tc(E 1.4 Tc(E = 0), where Tc(E = 0) is the critical temperature of the corresponding Ising model in thermodynamic equilibrium [715, 716, 1250, 1251]. On the other hand, Tc(E) decreases with E when the interparticle interactions are repulsive (i.e., J < 0); the ordering is altogether destroyed by sufficiently large E. However, there is no ordered structure at any nonzero temperature in the one-dimensional KLS model, irrespective of the sign of the interaction J.

Self-Driven Particles

The totally asymmetric simple exclusion process (TASEP) is probably the simplest model for a system of interacting self-driven particles. In this model, the configurations of the system are updated according to the following update rule: a particle moves forward with probability q if and only if the target site is empty. The particles are picked up either in parallel or in a random sequential manner for updating. The dynamics of a finite sample depends on the boundary conditions; most common boundary conditions are the periodic boundary conditions and open boundary conditions typically realized by coupling the system to particle reservoirs. Mathematically, TASEP is a special case of the KLS model. However, almost all the models of traffic can be regarded as extensions of the TASEP. Therefore, we present detailed discussion on the TASEP and briefly treat the KLS model as an extension of the TASEP in the following chapters.

1.8 PHASE TRANSITIONS, CRITICAL DYNAMICS, AND KINETICS OF PHASE ORDERING

1.8.1 Critical Dynamics: