Preface xv List of Figures xvii List of Tables xxxi List of Contributors xxxiii List of Abbreviations xxxvii 1 Introduction 1 1.1 Introduction 1 1.2 Brief History of Autonomous Vehicles 3 1.3 About the Book 4 2 Deep Learning for Autonomous Vehicles and Systems 9 2.1 Review of Deep Learning Models for AVS 10 2.1.1 Convolutional neural networks 10 2.1.2 Recurrent neural networks 12 2.1.3 Graph neural networks 14 2.1.4 Deep Q-networks 16 2.2 Tasks in Autonomous Vehicles and Systems 17 2.2.1 Perception 18 2.2.2 Mapping and localization 23 2.2.3 Path planing 25 2.2.4 Motion control learning 30 2.3 Future Trends and Challenges 32 2.3.1 Reliability 32 2.3.2 Computational and energy efficiency 33 2.3.3 Data sparsity 34 2.3.4 Human factors in AVS 35 2.4 Conclusions 36 3 Toward Autonomous Vehicles and Systems - Potentials and Challenges of Artificial Intelligence 49 3.1 Introduction 50 3.2 Modeling and Simulation 51 3.2.1 Vehicle models 51 3.2.1.1 Linear single-track model 51 3.2.1.2 Nonlinear single-track model 52 3.2.1.3 Linear roll and pitch model 53 3.2.2 Simulation setup 54 3.2.3 Data generation 59 3.2.4 Model training for the automated lane change 60 3.2.5 Model validation for the automated lane change 62 3.3 Results 65 3.3.1 Differences in vehicle model quality 65 3.3.2 Definition of the model boundaries 68 3.3.3 Development of a driver assistance system based on model quality 71 3.3.3.1 Data generation and model training 71 3.3.3.2 Evaluation of the trained support vector machines 73 3.3.3.3 Evaluation of the results 74 3.4 Conclusion 76 4 Autonomous Vehicle Steering through Convolutional and Recurrent Deep Learning 83 4.1 Introduction 84 4.2 Data Collection 86 4.2.1 Integrating software modules with the vehicle 88 4.3 Model Training 90 4.3.1 Deep learning fundamentals 90 4.3.2 Model structure 91 4.3.3 Training techniques 93 4.4 Results 96 4.4.1 Original CD (CNN+DNN) 96 4.4.2 Simple RNN 97 4.4.3 CD with new data and interchangeable pretrained networks 98 4.4.4 CDR with new data 101 4.5 Discussion and Future Work 104 5 2D and 3D Pose Estimation for Gesture Recognition in Deeplearning-driven Human-vehicle Leader-follower Systems 113 5.1 Introduction 114 5.1.1 Leader-follower background 114 5.1.2 Gesture recognition background 114 5.1.3 Previous work 115 5.2 Bipartite Mapping 117 5.3 Gesture Recognition 118 5.3.1 How to teach a machine 118 5.3.2 Gesture classification 119 5.3.3 Trials with a CNN 119 5.3.4 Modular data stream pipeline design 120 5.3.5 2D pose estimation 122 5.3.6 3D pose estimation 124 5.4 Integration of Models with the ACTor 1 Vehicle 125 5.4.1 ACTor 1 overview 125 5.4.2 ROS fundamentals 126 5.4.3 ROS node design and implementation 127 5.5 Testing Results 129 5.5.1 Results using the pipeline with 2D posenet 129 5.5.2 Bipartite mapping results 133 5.5.3 3D pose estimation results 133 5.6 Discussion 135 5.6.1 Summary 135 5.6.2 Future work 137 6 Adversarial Attacks and Defenses on Autonomous Vehicle Systems 143 6.1 Introduction 144 6.2 Attacks 147 6.2.1 White-box attacks 147 6.2.2 Black-box attacks 150 6.2.3 Patch attacks 152 6.2.4 Natural adversarial attacks 155 6.3 Defenses 156 6.3.1 Preprocessing defenses 156 6.3.2 Input-space sampling 159 6.3.3 Adversarial training 161 6.3.4 Certified defense 164 6.4 Conclusion 168 7 Multi-sensor Fusion for Multi-target Detection and Tracking 175 7.1 Introduction 175 7.2 Sensor System in Perception Module 178 7.2.1 Camera 178 7.2.2 LiDAR 179 7.2.3 MMW radar 180 7.2.4 4D MMW radar 181 7.2.5 Conclusion 182 7.3 Datasets 183 7.4 Multi-sensor Fusion Challenges and Solutions 186 7.4.1 Challenges 186 7.4.2 Solutions 187 7.4.2.1 Solutions for Challenges 1 and 2 188 7.4.2.1.1 Data Augmentation 188 7.4.2.1.2 Feature-level fusion 192 7.4.2.1.3 Object-level fusion 195 7.4.2.2 Sensor synchronization (solutions for challenge 4) 196 7.5 Multi-sensor Fusion in Detection and Tracking 197 7.5.1 Data processing representation 197 7.5.1.1 LiDAR point cloud representation 197 7.5.1.2 Radar data representation 198 7.5.2 3D object detection 199 7.5.2.1 Image-based 3D object detection 200 7.5.2.2 Point cloud-based 3D object detection 200 7.5.2.3 Multi-sensor-based 3D object detection 200 7.5.3 Multiple object tracking (MOT) 204 7.5.3.1 Vehicle side 204 7.5.3.2 Roadside 206 7.6 Conclusion 207 8 Teleoperation for Complementing and Enhancing Autonomous Vehicles 219 8.1 Background 220 8.1.1 Human intervention 220 8.1.2 Level 5 fully autonomous driving is still far away 220 8.2 Definition of Teleoperation 222 8.2.1 Teleoperation in connected mobility 222 8.2.2 Diverse forms of teleoperation 223 8.3 Importance of Teleoperation 224 8.3.1 Challenges of Fully Autonomous Driving 225 8.3.1.1 Failures in environment perception and recognition 225 8.3.1.2 Prolonged downtime due to the unfamiliar environmental interpretation 225 8.3.2 Unforeseen sensor malfunction and invalid sensor data 226 8.4 Teleoperation Legislation and Standards 227 8.4.1 Teleoperation legislation 227 8.4.1.1 Regulations in US states 227 8.4.1.2 Other countries' regulations 227 8.4.1.3 Illustration: Regulations on teleoperation in Germany 227 8.4.2 Teleoperation Standards 228 8.5 Technical Considerations and Improvement Techniques 229 8.5.1 Technical considerations 229 8.5.2 Improvement approaches 230 8.5.2.1 Perception enhancement for operators 230 8.5.2.2 Immersive virtual driving interface 231 8.5.2.3 Enhancement directions for control systems 231 8.5.2.4 Compensation for latency 232 8.6 Assessment Standards for Teleoperation Performance 232 8.7 Distinguished Participants in Automotive Teleoperation 233 8.7.1 Automotive company and startup 233 8.7.2 Companies offering teleoperation services 239 8.8 Teleoperation Case Studies 240 8.9 Open Problems and Potential Future Directions 242 8.9.1 Open problems in teleoperation 242 8.9.1.1 Bandwidth limitations 242 8.9.1.2 Fluctuating and excessive latency 243 8.9.1.3 Other open problems in teleoperation 244 8.9.2 Potential future directions 244 8.10 Conclusion 245 9 Hybrid Edge/Cloud Solutions for Supporting Autonomous Vehicles 255 9.1 Introduction 256 9.2 Related Work 257 9.3 System Architecture 258 9.4 Performance Evaluation 260 9.5 Conclusions 264 10 Game-theoretic Decision-making for Autonomous Driving Vehicles 269 10.1 Introduction 270 10.2 Problem Definition 273 10.2.1 Modeling scenarios for unsignalized intersection 273 10.2.1.1 Observation space 274 10.2.1.2 Action space 274 10.2.1.3 Reward function 275 10.2.2 Modeling scenarios for urban lane-change 275 10.2.2.1 Observation space 277 10.2.2.2 Action space 278 10.2.2.3 Reward function 278 10.2.3 End-to-end control scheme 279 10.3 Driver Interaction Model 279 10.3.1 level k reasoning 279 10.3.2 Dueling double deep Q-network with prioritized experience replay algorithm (D3QN PER) 280 10.3.3 Combining Level-K Reasoning with D3QN PER 282 10.3.4 Setting of Algorithm Parameters 284 10.4 Low Level Control 286 10.4.1 Trajectory generation module 286 10.4.2 Lateral control 286 10.4.3 Longitudinal control 287 10.5 Simulation and Experimental Results 287 10.5.1 Unsignalized intersection case 287 10.5.1.1 Comparison of different algorithms 287 10.5.1.2 Simulation results 288 10.5.1.3 Hardware Implementation 290 10.5.2 Urban lane-change case 292 10.5.2.1 Compare the trained policy with human drivers 292 10.6 Conclusions 294 10.6.1 Appendix 295 11 State of Modeling and Simulation in Autonomous Vehicles 303 11.1 Introduction 303 11.2 Building a world model in simulation 306 11.2.1 Components of a scenario 307 11.2.1.1 Defining scene 307 11.2.1.2 Defining situation 311 11.2.1.3 Defining scenario 314 11.2.2 Scenario descriptions in frameworks 315 11.2.2.1 PEGASUS and related projects - Germany 315 11.2.2.2 SAKURA - Japan 319 11.2.2.3 Other relevant frameworks 322 11.3 Sensor simulation - Inputting data into AV pipeline 322 11.3.1 Camera 323 11.3.2 LiDAR 326 11.3.3 Radar 328 11.4 Challenges and opportunities in AV simulation 329 12 Path Following Control System Design and
Autonomous vehicles and systems : a technological and societal perspective
2023
1 online resource (200 pages).
Campusweiter Zugriff (Universität Hannover) - Vervielfältigungen (z.B. Kopien, Downloads) sind nur von einzelnen Kapiteln oder Seiten und nur zum eigenen wissenschaftlichen Gebrauch erlaubt. Keine Weitergabe an Dritte. Kein systematisches Downloaden durch Robots.
Buch
Elektronische Ressource
Englisch
DDC: | 629.04/6 |
Exploring the Technological and Social Implications of Autonomous Vehicles
Springer Verlag | 2025
|Autonomous Vehicles: Present Technological Traits and Scope for Future Innovation
Springer Verlag | 2021
|IMPLICATIONS OF AUTONOMOUS VEHICLES: A PLANNER'S PERSPECTIVE
Online Contents | 2016
|Autonomous Vehicles: A Perspective from the California Department of Motor Vehicles
Springer Verlag | 2014
|