Safe, Autonomous and Intelligent Vehicles.pdf

发布时间：2022-06-13 发布人：admin 分类：说明书资料大小：4.75M 资料格式：pdf 举报版权申诉

ecd50ced-1a9d-438d-8a37-8c1a4fd21c5a.pdf-第1页.png

第1页 / 共215页

ecd50ced-1a9d-438d-8a37-8c1a4fd21c5a.pdf-第2页.png

第2页 / 共215页

ecd50ced-1a9d-438d-8a37-8c1a4fd21c5a.pdf-第3页.png

第3页 / 共215页

ecd50ced-1a9d-438d-8a37-8c1a4fd21c5a.pdf-第4页.png

第4页 / 共215页

ecd50ced-1a9d-438d-8a37-8c1a4fd21c5a.pdf-第5页.png

第5页 / 共215页

ecd50ced-1a9d-438d-8a37-8c1a4fd21c5a.pdf-第6页.png

第6页 / 共215页

ecd50ced-1a9d-438d-8a37-8c1a4fd21c5a.pdf-第7页.png

第7页 / 共215页

ecd50ced-1a9d-438d-8a37-8c1a4fd21c5a.pdf-第8页.png

第8页 / 共215页

Preface

Contents

About the Editors

1 Introduction

2 Efficient Statistical Validation of Autonomous Driving Systems

2.1 Introduction

2.2 Background

2.2.1 Image Sensing

2.2.2 Image Processing

2.2.3 Visual Perception

2.3 Test Data Generation

2.3.1 Temperature Variation

2.3.2 Circuit Aging

2.3.3 Corner Case Generation

2.3.4 Numerical Experiments

2.3.4.1 Experimental Setup

2.3.4.2 Temperature Variation

2.3.4.3 Circuit Aging

2.4 Subset Simulation

2.4.1 Mathematical Formulation

2.4.2 Random Sampling

2.4.3 Summary

2.4.4 Numerical Experiments

2.4.4.1 Experimental Setup

2.4.4.2 Experimental Results

2.5 Conclusions

References

3 Cyberattack-Resilient Hybrid Controller Design with Application to UAS

3.1 Introduction

3.2 Problem Formulation

3.2.1 System and Cyberattack Models

3.2.2 Cyberattack Mitigation Problem

3.3 Hybrid Controller Design

3.4 Analytical Performance Verification

3.5 Extension to Infinite Time Horizon: Receding Horizon Controller

3.6 Illustrative Example

3.6.1 H2 Optimal Controller

3.6.2 H∞ Optimal Controller

3.6.3 UAS Model

3.6.4 Simulation Results

3.7 Conclusions

References

4 Control and Safety of Autonomous Vehicles with Learning-Enabled Components

4.1 Hamilton–Jacobi Reachability

4.1.1 Backward Reachable Set (BRS)

4.1.2 Application: Provably Safe Multi-Vehicle Trajectory Planning

4.1.3 Limitations of HJ Reachability

4.2 Learning-Based Model Refinement

4.2.1 Function Approximator-Based Model Learning

4.2.2 Goal-Driven Model Learning

4.3 Safety Analysis of Learned Models

4.3.1 Safety During Model Learning

4.3.2 Model Validation Before Deployment

4.4 Learning in Partially Observable environments

References

5 Adaptive Stress Testing of Safety-Critical Systems

5.1 Introduction

5.2 Related Work

5.3 Background

5.3.1 Definitions

5.3.2 Sequential Decision Process

5.3.3 Monte Carlo Tree Search

5.4 Adaptive Stress Testing

5.4.1 Full Observability

5.4.2 Partial Observability

5.5 Aircraft Collision Avoidance Application

5.5.1 Experimental Setup

5.5.2 Results

5.5.3 Performance Comparison

5.6 Conclusion

References

6 Provably-Correct Compositional Synthesis of VehicleSafety Systems

6.1 Introduction

6.2 Autonomous Driving Functions

6.2.1 Adaptive Cruise Control

6.2.2 Lane Keeping

6.2.3 Challenges in Composition

6.3 Composition of Invariant Sets Via Contracts

6.3.1 Contract Realizability Problem

6.3.2 Contract Refinement Heuristic

6.4 Contract Realizability Via Polyhedral Controlled-Invariant Sets

6.4.1 Computation of Polyhedral Controlled-Invariant Sets

6.4.2 Over-Approximation of Nonlinear Parametrizations

6.4.3 Removal of Nonlinearities via Convexification

6.4.3.1 Convex-Hull Computation With Monotone Functions

6.4.3.2 Convex-Hull Computation With Convex Projections

6.5 Design Flow for the Case Study

6.5.1 Constraints

6.5.2 Contracts

6.5.3 Realizability of LK Contract

6.5.4 Realizability of ACC Contract

6.5.5 Low-Fidelity Simulation Results

6.5.6 CarSim Simulation Results

6.6 Implementation in Mcity

6.7 Conclusions

References

7 Reachable Set Estimation and Verification for Neural Network Models of Nonlinear Dynamic Systems

7.1 Introduction

7.2 Neural Network Models of Nonlinear Dynamic Systems

7.3 Problem Formulation

7.4 Reachable Set Estimation for MLPs

7.5 Reachable Set Estimation for NARMA Models

7.6 Magnetic Levitation Systems (Maglev)

7.6.1 Brief Introduction

7.6.2 Neural Network Model

7.6.3 Reachable Set Estimation

7.7 Conclusions

References

8 Adaptation of Human Licensing Examinations to the Certification of Autonomous Systems

8.1 Introduction

8.1.1 Driving Licensing Exams

8.1.2 Aviation Licensing Exams

8.2 SRKE Taxonomy

8.3 Implications of Human Licensing on Autonomous Vehicle Certification

8.3.1 Vision Tests for AVs?

8.3.2 Knowledge Tests and Checkrides for AVs?

8.3.3 Graduated Licensing

8.3.4 Certifying Machine Learning Algorithms Is Unprecedented

8.4 Conclusion

References

9 Model-Based Software Synthesis for Safety-Critical Cyber-Physical Systems

9.1 Software Challenges in Safety-Critical Cyber-Physical Systems

9.2 Model-Based Software Synthesis Flow

9.3 Holistic Timing-Driven Synthesis

9.4 Multi-Objective Optimization

9.4.1 Fault-Tolerance

9.4.2 Security

9.5 Cross-Layer Codesign

9.6 Conclusion

References

10 Compositional Verification for Autonomous Systems with Deep Learning Components

10.1 Introduction

10.2 Compositional Verification

10.3 Analysis for Deep Neural Network Components

10.3.1 ACAS Xu Case Study

10.4 Example

10.4.1 Run-Time Monitoring and Control

10.5 Conclusion

References

Index

Unmanned System Technologies Huafeng Yu Xin Li Richard M. Murray S. Ramesh Claire J. Tomlin Editors Safe, Autonomous and Intelligent Vehicles

Unmanned System Technologies

Springer’s Unmanned Systems Technologies (UST) book series publishes the latest developments in unmanned vehicles and platforms in a timely manner, with the highest of quality, and written and edited by leaders in the ﬁeld. The aim is to provide an effective platform to global researchers in the ﬁeld to exchange their research ﬁndings and ideas. The series covers all the main branches of unmanned systems and technologies, both theoretical and applied, including but not limited to: Unmanned aerial vehicles, unmanned ground vehicles and unmanned ships, and all unmanned systems related research in: Robotics Design Artiﬁcial Intelligence Guidance, Navigation and Control Signal Processing Circuit and Systems Mechatronics Big Data Advanced Materials and Engineering Intelligent Computing and Communication The publication types of the series are monographs, professional books, graduate textbooks, and edited volumes. More information about this series at http://www.springer.com/series/15608

Huafeng Yu Xin Li Richard M. Murray S. Ramesh Claire J. Tomlin Editors Safe, Autonomous and Intelligent Vehicles 123

Editors Huafeng Yu Boeing Research and Technology Huntsville, AL, USA Richard M. Murray California Institute of Technology Pasadena, CA, USA Claire J. Tomlin University of California Berkeley, CA, USA Xin Li Duke University Durham, NC, USA S. Ramesh General Motors R&D Warren, MI, USA ISSN 2523-3734 Unmanned System Technologies ISBN 978-3-319-97300-5 https://doi.org/10.1007/978-3-319-97301-2 ISSN 2523-3742 (electronic) ISBN 978-3-319-97301-2 (eBook) Library of Congress Control Number: 2018959861 © Springer Nature Switzerland AG 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, speciﬁcally the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microﬁlms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a speciﬁc statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional afﬁliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface An autonomous and intelligent system generally refers to a system that can automatically sense and adapt to dynamically varying environment. The term broadly covers numerous emerging and critical applications including self-driving vehicles, unmanned aircraft systems, and autonomous ships. A broad application of modern artiﬁcial intelligence and machine learning technologies is featured in these systems. The design, modeling, veriﬁcation, and validation of today’s autonomous and intelligent systems have become increasingly challenging with growing functional complexity in scale and features, the integration of new artiﬁcial intelligence and machine learning technologies, the adoption of more distributed and net- worked architectural platforms, and stringent demands on various design constraints imposed by performance, fault tolerance, reliability, extensibility, and security. The aforementioned trend on growing complexity presents tremendous design and validation challenges to safety assurance and certiﬁcation and calls for an immediate attention to this emerging area for developing radically new methodologies and practices to address the grand challenges, as well as enormous opportunities that have been rarely explored in the past. Over the past several years, a large number of academic articles, technical reports, and industrial whitepapers have been published in this area. However, due to the highly interdisciplinary nature of the area, they are often independently reported across diverse technical communities like veriﬁcation and validation, artiﬁcial intelligence, signal processing, system control, computer vision, and circuit design. Recent research and development in these areas has advanced to the point where an organized, integrated account seamlessly integrating the state of the art is immediately needed. This will help in comparing a large body of techniques in the literature and clarifying their trade-offs in terms of performance, cost, utility, etc. For this reason, there is increasing demand to report the state-of-the-art advances recently made by both academic and industrial researchers closely collaborating together in this area. v

vi Preface This book aims to answer this demand and to cover the important aspects of autonomous and intelligent systems, including perception, decision making, and control. It also covers the important application domains of these systems such as automobile and aerospace and, most importantly, how to deﬁne and validate the safety requirements as well as the designed systems, in particular machine learning enabled systems. To achieve these goals, rigorous veriﬁcation and validation methods are developed to address different challenges, based on formal methods, compositional synthesis, machine learning, adaptive stress testing, statistical validation, model-based design, and cyber resilience. The main objective of this book is to present the major challenges related to safety of next-generation machine learning enabled autonomous and intelligent systems with growing complexity and new applications, discuss new design and val- idation methodologies to address these safety issues, and offer sufﬁcient technical background to facilitate more academic and industrial researchers to collaboratively contribute to this emerging and promising area. We anticipate that this book will provide the knowledge and background for the recent research and development and, more importantly, bring together multiple communities for interdisciplinary cross-culture interaction and set the stage for future growth in the ﬁeld. Huntsville, AL, USA Durham, NC, USA Pasadena, CA, USA Warren, MI, USA Berkeley, CA, USA Huafeng Yu Xin Li Richard M. Murray S. Ramesh Claire J. Tomlin

Contents 1 2 3 4 5 6 7 8 Introduction ................................................................. Huafeng Yu, Xin Li, Richard M. Murray, S. Ramesh, and Claire J. Tomlin Efﬁcient Statistical Validation of Autonomous Driving Systems ...... Handi Yu, Weijing Shi, Mohamed Baker Alawieh, Changhao Yan, Xuan Zeng, Xin Li, and Huafeng Yu Cyberattack-Resilient Hybrid Controller Design with Application to UAS ................................................... Cheolhyeon Kwon and Inseok Hwang Control and Safety of Autonomous Vehicles with Learning-Enabled Components ..................................... Somil Bansal and Claire J. Tomlin Adaptive Stress Testing of Safety-Critical Systems ..................... Ritchie Lee, Ole J. Mengshoel, and Mykel J. Kochenderfer Provably-Correct Compositional Synthesis of Vehicle Safety Systems .............................................................. Petter Nilsson and Necmiye Ozay 1 5 33 57 77 97 Reachable Set Estimation and Veriﬁcation for Neural Network Models of Nonlinear Dynamic Systems .................................. 123 Weiming Xiang, Diego Manzanas Lopez, Patrick Musau, and Taylor T. Johnson Adaptation of Human Licensing Examinations to the Certiﬁcation of Autonomous Systems ............................. 145 M. L. Cummings vii

分享到：

赞收藏

资料库

Safe, Autonomous and Intelligent Vehicles.pdf

相关推荐

课程资源

热门标签

最新资料