Modelling Driver Behaviour in Automotive Environments (eBook)

Carlo Cacciabue is an internationally-renowned expert on safety-critical systems and accident investigation, and an experienced consultant. He is a senior scientist in the Institute for the Protection and Security of the Citizen, at the Joint Research Centre of the European Commission, Ispra, Italy. He holds a PhD in Nuclear Engineering from the Politecnico di Milano (Italy) and retains a number of temporary academic positions in Italian and European universities. He is the author of two monographs and of several publications in journals and books relating to safety engineering and human-machine interaction in the domains of energy production and transportation. He is the principal Editor, together with E. Hollnagel, of Springer's International Journal of Cognition, Technology & Work.

Title Page 3
Copyright Page 4
Table of Contents 5
Editorial 8
List of Contributors 12
I International Projects and Actions on Driver Modelling 15
1 Modelling Driver Behaviour in European Union and International Projects 16
1.1 Introduction 16
1.2 Evaluation of Driver Behaviour Models 17
1.2.1 Michon's Hierarchical Control Model 17
1.2.2 The GADGET-Matrix: Integrating Hierarchical Control Models and Motivational Models of Driver Behaviour 18
1.2.3 DRIVABILITY Model 19
1.3 Driver Behaviour Adaptation Models and Their Relation to ADAS 22
1.3.1 Automaticity 24
1.3.2 Locus of Control 24
1.3.3 Risk Homeostasis 25
1.3.4 Risk Compensation 26
1.3.5 Threat Avoidance 26
1.3.6 Utility Maximisation 27
1.3.7 Behavioural Adaptation Formula 27
1.4 Use of Driver Behaviour Models in EU and International Projects 28
1.4.1 Driver Models Use for Driver Training and Assessment 28
1.4.2 Evaluation ofDriver Models' Use for Safety Aids 28
1.4.2.1 Use of Seat Belts 28
1.4.2.2 Use of Motorcycle Helmet 30
1.4.2.3 Studded Tyres 32
1.4.2.4 Antilock Braking Systems 32
1.4.3 Driver Models Use for ADAS Design and Impact Assessment 33
1.5 Conclusions 35
References 36
2 TRB Workshop on Driver Models: A Step Towards a Comprehensive Model of Driving? 39
2.1 Introduction 39
2.2 Workshop Presentation and Speakers' Contribution 40
2.2.1 Workshop Content 40
2.2.1.1 Driver Model Purpose and Application 40
2.2.1.2 Driver Model Architecture and Implementations 40
2.2.1.3 Calibration and Validation 41
2.2.2 Summaries ofthe Speakers' Contributions 42
2.2.2.1 In-Vehicle Information System - Jon Hankey 42
2.2.2.2 ACT-R Driver Model- Dario Salvucci 43
2.2.2.3 Optimal Control Model - Richard van der Horst 45
2.2.2.4 ACME 47
2.2.2.5 Fuzzy Logic Based Motorway Simulation 48
2.3 Synthesis of Presented Models 49
2.3.1 Understanding Models' Scope 49
2.3.2 Driver Model Toolbox 51
2.4 Towards a Comprehensive Model of Driving 52
2.5 Conclusions 53
References 55
3 Towards Monitoring and Modelling for Situation-Adaptive Driver Assist Systems 56
3.1 Introduction 56
3.2 Behaviour-Based Human Environment Creation Technology Project 57
3.2.1 Aims of the Project 57
3.2.2 Measurement of Driving Behaviour 58
3.2.3 Driving Behaviour Modelling 58
3.2.4 Detection of Non-Normative Behaviour 58
3.2.5 Estimation of Driver's State 59
3.2.5.1 Estimation of Driver 's Mental Tension 59
3.2.5.2 Estimation of Driver's Fatigue 60
3.3 Situation and Intention Recognition for Risk Finding and Avoidance Project 60
3.3.1 Aims of the Project 60
3.3.2 Adaptive Function Allocation Between Drivers and Automation 63
3.3.3 Decision Authority and the Levels of Automation 64
3.3.4 Model-Based Evaluation of Levels of Automation 65
3.3.4.1 Drivers' Psychological States and Their Transitions 66
3.3.4.2 Driver's Response to an Alert 66
3.3.4.3 Evaluation of Efficacy of Levels of Automation 67
3.4 Concluding Remarks 67
References 69
II Conceptual Framework and Modelling Architectures 71
4 A General Conceptual Framework for Modelling Behavioural Effects of Driver Support Functions 72
4.1 Introduction 72
4.2 Intended Application Areas and Requirements 73
4.2.1 Functional Characterisation of Driver Support Functions 73
4.2.2 Coherent Description ofExpected Behavioural Effects of Driver Support Functions 73
4.2.3 Conceptualising Relations Between Behavioural Effects and Road Safety 74
4.2.4 Specific Requirements 74
4.3 Existing Models of Driver Behaviour 74
4.3.1 Manual Control Models 74
4.3.2 Information Processing Models 75
4.3.3 Motivational Models 75
4.3.4 Safety Margins 76
4.3.5 Hierarchical Models 77
4.4 A Conceptual Framework 77
4.4.1 Driver Behaviour as Goal-Directed Activity 78
4.4.2 Dynamical Representation of Driver Behaviour 78
4.4.3 The Contextual Control Model (COCOM) 79
4.4.4 The Extended Control Model (ECOM) 81
4.5 Application 83
4.5.1 Characterising Driver Support Functions 83
4.5.1.1 Support for Tracking 83
4.5.1.2 Support for Regulating 84
4.5.1.3 Support for Monitoring 84
4.5.1.4 Support for Targeting 84
4.5.1.5 Non-Driving-Related Functions 85
4.5.1.6 Workload Management Functions 85
4.5.2 Characterising Behavioural Effects of Driver Support Functions 85
4.5.2.1 Behavioural Adaptation to Driving Support Functions 86
4.5.2.2 Effects of Multitasking While Driving 87
4.5.3 Driver Behaviour and Accident Risk 89
4.6 Discussion and Conclusions 91
References 92
5 Modelling the Driver in Control 96
5.1 Introduction 96
5.2 A Cognitive View of Driving 96
5.3 Human Abilities 97
5.4 Classifying Driver Behaviour Models 98
5.5 Hierarchical Control Models 98
5.6 Control Theory 100
5.7 Adaptive Control Models 102
5.8 Cognition in Control 104
5.9 Goals for Control 106
5.10 Time and Time Again 108
5.11 Multiple Layers of Control 109
5.12 Joystick Controlled Cars - An Example 111
5.13 Summary and Conclusion 112
References 113
6 From Driver Models to Modelling the Driver: What Do We Really Need to Know About the Driver? 116
6.1 Introduction 116
6.2 A Typology of Models 117
6.3 Descriptive Models 117
6.3.1 Task Models 117
6.3.2 Adaptive Control Models 118
6.3.3 Production Models 118
6.4 Motivational Models 120
6.5 Towards a Real-Time Model of the Driver 123
6.5.1 What Type of Model Is Required? 123
6.5.2 Grouping the Factors 124
6.5.3 A Proposed Structure 126
6.5.4 Verifying the Model 127
6.6 Developing an Online Model 128
6.7 Conclusions 130
References 130
III Learning and Behavioural Adaptation 132
7 Subject Testing for Evaluation of Driver Information Systems and Driver Assistance Systems - Learning Effects and Methodological Solutions 133
SUMMARY 133
7.1 Introduction 133
7.2 Methodological Issues 135
7.3 Experimental Examples 136
7.3.1 Evaluation of a Multimodal HMI 137
7.3.2 Destination Entry While Driving 139
7.3.3 Evaluation of Driver Assistance Systems 140
7.4 Solutions 141
7.5 Conclusions 142
References 143
8 Modelling Driver's Risk Taking Behaviour 145
8.1 Introduction 145
8.2 Expected Risk Reductions from New Technology on the Road 145
8.3 Behaviour When Driving with Supports 146
8.3.1 The Importance of Plain Old Ergonomics 146
8.3.2 The Loss of Potentially Useful Skills 146
8.3.3 Opportunities for New Errors 146
8.3.4 Problematic Transitions 147
8.3.5 Risk and Risk Perception: My Risk and Yours 147
8.4 Behavioural Adaptation 147
8.4.1 Direct Changes in Behaviour 147
8.4.2 A Word of Caution About Working with Risk Measures in Traffic Safety Studies 149
8.4.3 A Piece of Empirical Evidence from Seat Belt Accident Statistics 150
8.4.4 Higher-Order Forms ofAdaptation 151
8.5 The Link Between Behaviour and Risk 152
8.5.1 Average Speed, Speed Variability and Risk 152
8.5.2 Lane-Keeping Performance and Risk 152
8.5.3 Car-Following and Risk 153
8.6 Countermeasures Against Behavioural Adaptation 154
8.6.1 Should There Be Any? 154
8.6.2 Incentive Schemes and Their Expected Results 154
8.7 Conclusions 154
8.8 An Afterthought 154
References 155
9 Dealing with Behavioural Adaptations to Advanced Driver Support Systems 157
9.1 Introduction 157
9.2 'Behavioural Adaptation' in Road Safety Research 158
9.3 Behavioural Adaptation to Advanced Driver Support Systems 159
9.3.1 The Diversity of Behavioural Changes Studied and Observed 160
9.3.2 The Importance of the Situational Context and the Interactive Dimension of Driving 162
9.3.3 The Potential Differential Impact of Driver Support Systems 163
9.3.4 Learning to Drive with New Driver Support Systems 165
9.4 Behavioural Adaptation in the AIDE Project 167
References 168
IV Modelling Motivation and Psychological Mechanisms 172
10 Motivational Determinants of Control in the Driving Task 173
10.1 Introduction 173
10.2 Understanding Speed Choice 173
10.2.1 Behaviour Analysis 173
10.2.2 The Theory of Planned Behaviour 175
10.2.3 Risk Homeostasis Theory 177
10.2.4 The Task-Capability Interface Model 179
10.2.4.1 The Determination of Task-Difficulty Level: Task-Difficulty Homeostasis 182
10.2.4.2 The Representation of Task-Difficulty 185
10.2.5 The Somatic-Marker Hypothesis 187
10.2.5.1 Predictions and Speculations from the Somatic-Marker Hypothesis 189
10.3 Conclusions 191
References 193
11 Towards Understanding Motivational and Emotional Factors in Driver Behaviour: Comfort Through Satisficing 197
11.1 Introduction 197
11.2 Emotional Tension and 'Risk Monitor' 198
11.3 Safety Margins and Safety Zone 199
11.4 Available Time, Workload and Multilevel Task Control 201
11.5 Safety Margins, Affordances and Skills 204
11.6 Towards Unifying Emotional Conceptsin Routine Driving 206
11.6.1 Safety Margins - To Control and Survive 207
11.6.2 Vehicle/Road System - To Provide Smooth and Comfortable Travel 208
11.6.3 Rule Following - ToAvoid Sanctions 208
11.6.4 Good (or Expected) Progress of Trip -Mobility and Pace/Progress 209
11.7 Comfort Through Satisficing 209
11.8 'Go to the Road': Need of On-Road Research 211
References 212
12 Modelling Driver Behaviour on Basis of Emotions and Feelings: Intelligent Transport Systems and Behavioural Adaptations 216
12.1 Introduction 216
12.2 Defining Motivation 216
12.3 Motivational Aspects in Driver Behaviour Models 217
12.4 Behavioural Adaptation and Risk Compensation 218
12.5 Wilde's Risk Homeostasis Theory (RHT) 219
12.5.1 Target Risk or Target Feeling? 222
12.6 Effects of ABS: An Illustrative Example of ITS 223
12.7 Issues Raised by the Example of ABS: The Relevance for ITS 226
12.8 Adaptation - Mismatch Between Technology and Human Capability 227
12.9 ITS Technology May Enhance As Well As Reduce the Window of Opportunities 228
12.10 Damasio and the Somatic Marker Hypothesis 229
12.11 The Monitor Model 232
12.12 The Monitor Model and Prediction of Effects of ITS 235
12.13 Summary and Conclusions 237
References 238
V Modelling Risk and Errors 241
13 Time-Related Measures for Modelling Risk in Driver Behaviour 242
13.1 Introduction 242
13.2 The Driving Task 243
13.3 Lateral Control 245
13.3.1 Time-to-Line Crossing (TLC) 245
13.3.2 Lateral Distance When Passing 246
13.4 Longitudinal Control 247
13.4.1 Time-to-Collision (ITC) 247
13.4.2 Time-to-Intersection (TTl) 255
13.4.3 Time-to-Stop-Line (ITS) 256
13.5 Conclusions 257
References 257
14 Situation Awareness and Driving: A Cognitive Model 260
14.1 Introduction 260
14.2 Situation Awareness 260
14.2.1 An Algorithmic Description of Situation Awareness 261
14.2.1.1 The Construction of the Situation Model: Comprehending the Situation 262
14.2.1.2 Selection of Actions and the Control of Behaviour 264
14.3 Errors and the Comprehension Based-Model of Situation Awareness 265
14.4 Situation Awareness and In-Vehicle Information System Tasks 267
14.4.1 A Measurement Procedure: Context-Dependent Choice Reaction Task 267
14.4.2 Evaluation of the Context-Dependent Choice Reaction Task 269
14.5 Conclusions 270
References 271
15 Driver Error and Crashes 273
15.1 Slips, Lapses and Mistakes 273
15.2 Errors and Violations 274
15.3 The Manchester Driver Behaviour Questionnaire 275
15.4 The DBQ and Road Traffic Accidents 275
15.5 Aggressive Violations 278
15.6 Anger-Provoking Situations 279
15.7 Conclusions 280
References 280
VI Control Theory Models of Driver Behaviour 282
16 Control Theory Models of the Driver 283
16.1 Introduction 283
16.2 Modelling Human Controlling Behaviour 283
16.2.1 The Tustin-Model: Linear Part + Remnant 283
16.2.2 Laboratory Research, Stochastic Input, Quasi-Linear Model 285
16.2.3 A Holistic Approach: The Crossover Model 286
16.2.4 Nonlinear Approaches: Improved Reproduction of Measured Behaviour 287
16.3 Driver Models for Vehicle Design 289
16.4 Summary and Future Prospects 295
References 296
17 Review of Control Theory Models for Directional and Speed Control 299
17.1 Introduction 299
17.2 Basic Crossover Model of the Human Operator 300
17.3 Model for Driver Steering Control 302
17.3.1 Equivalent Single-Loop System for Steering Control 304
17.4 Model for Speed Control with Accelerator Pedal 305
17.5 Experimental Data 308
17.5.1 Driving Simulator Measurements 308
17.5.1.1 Steering Control 309
17.5.1.2 Speed Control 310
17.5.2 Actual Vehicle Measurements 312
17.6 Example Directional Control Application 312
17.7 Discussion 316
References 316
VII Simulation of Driver Behaviour 318
18 Cognitive Modelling and Computational Simulation of Drivers Mental Activities 319
18.1 Introduction: A Brief Historical Overview on Driver Modelling 319
18.2 COSMODRIVE Model 321
18.2.1 Cognitive Architecture ofCOSMODRIVE 321
18.2.2 The Tactical Module 323
18.2.2.1 Driving Frames: A Framework for Modelling Mental Models 324
18.2.2 .2 Architecture of the Tactical Module 328
18.2.2.3 The Blackboards of the Tactical Module 329
18.2.2.4 The Knowledge Bases (KB) of the Tactical Module 330
18.2.2.5 The Cognitive Processes of the Tactical Module 331
18.2.2.5.1 Categorisation 332
18.2.2.5.2 The Place Recognition Process 332
18.2.2.5.3 The Tactical Representations Generator Process 332
18.2.2.5.4 The Anticipation Process 334
18.2.2.5.5 The Decision Process 335
18.3 Methodology to Study Driver's Situation Awareness 336
18.3.1 Main Hypothesis 336
18.3.2 Methodology 337
18.3.3 Main Results 337
18.3.4 Discussion and Conclusion Concerning Experimental Study of Drivers Situation Awareness 340
18.4 Some Experimental Results Simulation with Cosmodrive 341
18.5 Conclusion and Perspectives: From Behaviours to Mental Model 343
References 345
19 Simple Simulation of Driver Performance for Prediction and Design Analysis 348
19.1 Introduction 348
19.1.1 Modelling Human Behaviour in Modern Technology 348
19.1.2 Modelling Drivers in the Automotive Context 349
19.1.3 Use and Applications ofDriver Models 351
19.1.4 Content ofthe Paper 352
19.2 Simple Simulation of Driver Behaviour 352
19.2.1 Paradigm of Reference 352
19.2.2 Simulation Approach for Normative Behaviour 353
19.2.2.1 Task Analysis 353
19.2.2.2 Dynamic Logical Simulation of Tasks 354
19.2.3 Algorithms for Cognition, Behavioural Adaptation and Errors 356
19.2.3.1 Normative Driver Behaviour 358
19.2.3.2 Descriptive Driver Behaviour 359
19.2.3.3 Parameters and Measurable Variables 361
19.2.3.3.1 Task Demand 362
19.2.3.3.2 Driver State 363
19.2.3.3.3 Situation Awareness 365
19.2.3.4 Intentions, Decision Making and Human Error 367
19.2.3.4.1 Intentions and Decision Making 368
19.2.3.4.2 Error Generation 369
19.2.4 Simulation of Control Actions 370
19.2.4.1 Normal Driving 371
19.2.4.2 Error in Control Actions 373
19.3 Sample Cases of Predictive DVE Interactions 375
19.3.1 Case 1 375
19.3.2 Case 2 377
19.4 Conclusions 378
References 378
VIII Simulation of Traffic and Real Situations 380
20 Real-Time Traffic and Environment Risk Estimation for the Optimisation of Human-Machine Interaction 381
20.1 Introduction 381
20.2 The AWAKE Use Case - Adaptation of a Driver Hypovigilance Warning System 382
20.2.1 AWAKE System Overview 382
20.2.2 Traffic Risk Estimation in AWAKE System 383
20.2.3 The Scenario-Assessment Unit 384
20.2.4 The Warning Strategies Unit 384
20.2.5 The Risk-Level Assessment Unit 385
20.3 The AIDE Use Case - Optimisation of the In-Vehicle Human-Machine Interaction 386
20.3.1 Overview 387
20.3.2 Architecture 388
20.3.2.1 Relevance to the AIDE Use Cases 388
20.3.2.2 Description of Environment 389
20.3.3 Algorithm for Risk Assessment 390
20.3.3.1 Rule-Based System Employed for TERA Algorithms 390
20.3.3.2 Main Traffic Risk Condition Detection 392
20.3.3.2.1 Risk of Frontal/(Lateral) Collision 393
20.3.3.2.2 Criteria of Assigning the Level of Risk 393
20.3.3.2.3 Risk of Lane/Road Departure 394
20.3.3.2.4 Risk of Approaching a Dangerous Curve Too Fast 395
20.3.4 Algorithmfor Estimating the Intention of the Driver 395
20.3.5 TERA Implementation 397
20.4 Conclusions 399
References 400
21 Present and Future of Simulation Traffic Models 402
21.1 Introduction 402
21.2 Traffic Simulator 403
21.2.1 General Overview: A Survey of Road Traffic Simulations 403
21.2.2 Types of Simulator 405
21.2.3 Case Studies of Traffic Simulator 407
21.2.4 Vehicle Model Properties 409
21.2.4.1 Perception Topics 411
21.2.4.2 Cognition Topics 412
21.2.4.3 Actuation/Control Topics 413
21.2.4.4 Implementation of Vehicle Model 413
21.2.5 Two Examples of Applications with Traffic Simulator 414
21.2.5.1 The University of Michigan Microscopic Traffic Simulator 415
21.2.5.2 The MECTRON-HMI Group at University of Modena and Reggio Emilia Driving Simulator used in Human factors and Human-Machine Interfaces Studies. 416
21.2.6 Integration of Driver, Vehicle and Environment in a Closed-Loop System: The AIDE Project 419
21.2.6.1 General DVE Architecture 420
21.2.6.2 Time Frame for DVE Model 421
21.3 Conclusions and Further Steps 422
21.3.1 Towards a Multi-Agent Approach 423
21.3.2 New Developments and Prospective 423
21.3.3 Open Points and Future Steps 424
References 426
Index 430

"V Modelling Risk and Errors 13 Time-Related Measures for Modelling Risk in Driver Behaviour (p. 234-236)

RICHARD VAN DER HORST

13.1 Introduction

Accident statistics have an important general safety monitoring function and form a basis for detecting specific traffic safety problems. However, the resulting information is inadequate for analysing and diagnosing, defining remedial measures and evaluating their effects. Systematic observations of driver behaviour, combined with knowledge of human information-processing capabilities and limitations, offer wider perspectives in understanding the causes of safety problems and modelling driver behaviour in both normal and critical situations.

Renewed interest results from the need to develop, test, assess and evaluate driver support systems in terms of drivers behaviour, performance and acceptance. The processes that result in near-accidents or traffic conflicts have much in common with the processes preceding actual collisions (Hyden, 1987); only the final outcome is different. The frequency of traffic conflicts is relatively high, and they offer a rich information source on causal relationships since the preceding process can be systematically observed. In this approach, traffic situations are ranked along a continuum ranging from normal situations, via conflicts to actual collisions.

A pyramidal representation of this continuum was introduced by Hyden (1987), clearly visualising the relative rate of occurrence of the different events (Fig. 13.1). The analysis of driver behaviour in critical encounters may not only offer a better understanding of the processes that ultimately result in accidents, but, perhaps even more efficient in the long run, also provide us with knowledge on drivers abilities of turning a critical situation into a controllable one.

A general conceptual description of the driving task as commonly used in traffic psychonomics, with time-to-line crossing (TLC) and time-to-collision (TTC) as a measure for describing the lateral and longitudinal driving task will be used to distinguish normal from critical behaviour. That may serve as realistic criterion settings for in-car warning systems such as forward collision warning and intersection collision avoidance warning systems.

13.2 The Driving Task

In the literature the task analy sis for driving a car is well documented. A frequently used conceptual model of the driving task consists of three hierarchically ordered levels, navigation, guidance and control (Allen et aI., 1971). In other publications these levels are also referred to as strategic, manoeuvring and control. Tasks at the navigat ion level refer to the activities related to planning and executing a trip from origin to destination. The need for processing information only occurs occasionally, with intervals ranging from a few minutes to hour s. The guidance level refer s to tasks deal ing with the interaction with both environment (roadway, traffic signs , traffic signals) and other road users. Activity is required rather frequently with intervals of a few seconds to a few minutes."