Autonomous Road Vehicle Path Planning and Tracking Control E-Book

Levent Guvenc

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Serie: Wiley-IEEE Press Book Series on Control Systems Theory and Applications
• Sprache: Englisch

Seitenzahl: 435

Veröffentlichungsjahr: 2021

Beschreibung

Discover the latest research in path planning and robust path tracking control In Autonomous Road Vehicle Path Planning and Tracking Control, a team of distinguished researchers delivers a practical and insightful exploration of how to design robust path tracking control. The authors include easy to understand concepts that are immediately applicable to the work of practicing control engineers and graduate students working in autonomous driving applications. Controller parameters are presented graphically, and regions of guaranteed performance are simple to visualize and understand. The book discusses the limits of performance, as well as hardware-in-the-loop simulation and experimental results that are implementable in real-time. Concepts of collision and avoidance are explained within the same framework and a strong focus on the robustness of the introduced tracking controllers is maintained throughout. In addition to a continuous treatment of complex planning and control in one relevant application, the Autonomous Road Vehicle Path Planning and Tracking Control includes: * A thorough introduction to path planning and robust path tracking control for autonomous road vehicles, as well as a literature review with key papers and recent developments in the area * Comprehensive explorations of vehicle, path, and path tracking models, model-in-the-loop simulation models, and hardware-in-the-loop models * Practical discussions of path generation and path modeling available in current literature * In-depth examinations of collision free path planning and collision avoidance Perfect for advanced undergraduate and graduate students with an interest in autonomous vehicles, Autonomous Road Vehicle Path Planning and Tracking Control is also an indispensable reference for practicing engineers working in autonomous driving technologies and the mobility groups and sections of automotive OEMs.

Leseprobe

Table of Contents

Cover

Title Page

Copyright

Dedication

About the Authors

Preface

List of Abbreviations

1 Introduction

1.1 Motivation and Introduction

1.2 History of Automated Driving

1.3 ADAS to Autonomous Driving

1.4 Autonomous Driving Architectures

1.5 Cybersecurity Considerations

1.6 Organization and Scope of the Book

1.7 Chapter Summary and Concluding Remarks

References

2 Vehicle, Path, and Path Tracking Models

2.1 Tire Force Model

2.2 Vehicle Longitudinal Dynamics Model

2.3 Vehicle Lateral Dynamics Model

2.4 Path Model

2.5 Pure Pursuit: Geometry‐Based Low‐Speed Path Tracking

2.6 Stanley Method for Path Tracking

2.7 Path Tracking in Reverse Driving and Parking

2.8 Chapter Summary and Concluding Remarks

References

3 Simulation, Experimentation, and Estimation Overview

3.1 Introduction to the Simulation‐Based Development and Evaluation Process

3.2 Model‐in‐the‐Loop Simulation

3.3 Virtual Environments Used in Simulation

3.4 Hardware‐in‐the‐Loop Simulation

3.5 Experimental Vehicle Testbeds

3.6 Estimation

3.7 Chapter Summary and Concluding Remarks

References

4 Path Description and Generation

4.1 Introduction

4.2 Discrete Waypoint Representation

4.3 Parametric Path Description

4.4 Tracking Error Calculation

4.5 Chapter Summary and Concluding Remarks

References

5 Collision Free Path Planning

5.1 Introduction

5.2 Elastic Band Method

5.3 Path Planning with Minimum Curvature Variation

5.4 Model‐Based Trajectory Planning

5.5 Chapter Summary and Concluding Remarks

References

6 Path‐Tracking Model Regulation

6.1 Introduction

6.2 DOB Design and Frequency Response Analysis

6.3 Q Filter Design

6.4 Time Delay Performance

6.5 Chapter Summary and Concluding Remarks

References

7 Robust Path Tracking Control

7.1 Introduction

7.2 Model Predictive Control for Path Following

7.3 Design Methodology for Robust Gain‐Scheduling Law

7.4 Robust Gain‐Scheduling Application to Path‐Tracking Control

7.5 Add‐on Vehicle Stability Control for Autonomous Driving

7.6 Chapter Summary and Concluding Remarks

References

8 Summary and Conclusions

8.1 Summary

8.2 Conclusions

Index

End User License Agreement

List of Tables

Chapter 4

Table 4.1 Discrete waypoint processing procedures.

Chapter 5

Table 5.1 Advantages and disadvantages of some path planning methods.

Table 5.2 Algorithm for modification of vehicle path nodes.

Table 5.3 RMS values for tracking error.

Table 5.4 Time cost comparison.

Chapter 7

Table 7.1 Comparisons of tracking errors during experiments.

List of Illustrations

Chapter 1

Figure 1.1 Categories of autonomous driving according to SAE.

Figure 1.2 Illustration of radio‐controlled driverless car of 1920s.

Figure 1.3 First university based self‐driving study by Ohio State Universit...

Figure 1.4 Analog steering and speed control of 1960s Ohio State University ...

Figure 1.5 Joystick for drive‐by‐wire interface of Ohio State University sel...

Chapter 2

Figure 2.1 Geometry of the wheel and tire force model parameters.

Figure 2.2 Unsymmetrical pressure distribution in a rolling tire and rolling...

Figure 2.3 Deformation in contact zone due to lateral force and self‐alignin...

Figure 2.4 Longitudinal tire force as a function of slip.

Figure 2.5 m‐File for generating longitudinal magic tire force.

Figure 2.6 Lateral tire force as a function of sideslip angle.

Figure 2.7 Camber thrust.

Figure 2.8 m‐File for generating lateral magic tire force.

Figure 2.9 Self‐aligning moment as a function of sideslip angle.

Figure 2.10 Friction circle (Kamm Circle).

Figure 2.11 Simple Dugoff tire model forces (a) as a function of slip ratio ...

Figure 2.12 MNC tire forces (a) as a function of slip ratio (b) as a functio...

Figure 2.13 m‐File for the MNC tire forces function.

Figure 2.14 Longitudinal road load.

Figure 2.15 Vehicle longitudinal dynamics block diagram.

Figure 2.16 Free body diagram of drive wheel.

Figure 2.17 Block diagram of wheel dynamics.

Figure 2.18 Longitudinal dynamics combined with one wheel model.

Figure 2.19 Geometry of high‐speed cornering.

Figure 2.20 Geometry of low‐speed cornering.

Figure 2.21 Illustration of understeer and oversteer.

Figure 2.22 Geometry of single‐track vehicle model.

Figure 2.23 Block diagram of nonlinear single‐track vehicle model.

Figure 2.24 Augmented single‐track vehicle model.

Figure 2.25 Yaw gain versus speed for oversteer and understeer vehicles.

Figure 2.26 Desired path and tracking error.

Figure 2.27 Path‐tracking model detailed geometry.

Figure 2.28 Desired path translated to pass through vehicle center of mass....

Figure 2.29 Minimum distance between vehicle and path.

Figure 2.30 Angular orientation of path.

Figure 2.31 Pure pursuit steering.

Figure 2.32 Stanley method.

Figure 2.33 Path tracking in reverse driving.

Chapter 3

Figure 3.1 Illustration of DIL simulation.

Figure 3.2 V diagram for model, XIL, and road testing of autonomous driving ...

Figure 3.3 Linearized vehicle path following model.

Figure 3.4 Nonlinear vehicle path following model.

Figure 3.5 Generic E‐Class sedan subsystem properties in CarSim.

Figure 3.6 CarSim simulation environment with traffic, environment, and sens...

Figure 3.7 OpenStreetMap editing and exporting.

Figure 3.8 Some roads and intersections in Linden Residential Environment cr...

Figure 3.9 Pipelined approach for importing an OSM map into the Unity Engine...

Figure 3.10 OpenStreetMap (a) and corresponding unedited, imported map in Un...

Figure 3.11 Three‐dimensional model of the simulation environment where road...

Figure 3.12 LGSVL simulator autonomous vehicle (a) controlled by Autoware (b...

Figure 3.13 Pipelined approach for importing map data into the Unreal Engine...

Figure 3.14 Local features of the mesh of the Linden Residential Area.

Figure 3.15 Unity‐based LGSVL simulator emulation of 3D lidar and camera sen...

Figure 3.16 Three‐dimensional lidar point cloud emulation reflected on envir...

Figure 3.17 Depth, RGB, and semantic segmentation cameras in Unity‐based LGS...

Figure 3.18 Radar sensor detections with simulation of traffic using NPC veh...

Figure 3.19 CARLA AV simulation in the Linden Residential Area.

Figure 3.20 Depth camera display in the CARLA simulator.

Figure 3.21 3D lidar scan in the CARLA simulator.

Figure 3.22 HIL simulator components.

Figure 3.23 Information flow between HIL simulator components.

Figure 3.24 Unified AV architecture.

Figure 3.25 Experimental vehicle using unified architecture.

Figure 3.26 Unified multiobjective parameter space approach for trajectory f...

Figure 3.27 Uncertainty boxes of operating conditions for robustness analysi...

Figure 3.28 Vehicle mass and speed uncertainty regions for two categories of...

Figure 3.29 Unified library of AV functions and sensors.

Figure 3.30 Dash EV experimental vehicle using unified architecture.

Figure 3.31 Normalized longitudinal force as a function of slip computed usi...

Figure 3.32 Estimation diagram of the slipslope method.

Figure 3.33 Experiment road condition.

Figure 3.34 Vehicle longitudinal (a) and lateral (b) speeds estimated with t...

Figure 3.35 Measured wheel rotational speeds during the experiment.

Figure 3.36 Scatter plot of the rear tire longitudinal speed and its rotatio...

Figure 3.37 Effective tire radius estimated for the rear two wheels.

Figure 3.38 Scatter plot of the normalized longitudinal force versus the ave...

Figure 3.39 Estimated value of the slipslope updated using RLS with respect ...

Figure 3.40 Road surfaces designed in simulation environment.

Figure 3.41 Vehicle longitudinal motion and estimated slipslope as well as r...

Chapter 4

Figure 4.1 Comparison of: (a) positions of the original, adjusted, and optim...

Figure 4.2 Illustration of parametric path description.

Figure 4.3 Cornu spiral in the

x

–

y

–

s

space and its plane projections, at

κ0 

...

Figure 4.4 Clothoid tentacles generated at different initial condition.

Figure 4.5 The forming of Bezier curve with De Casteljau's algorithm [11].

Figure 4.6 Bezier curve generation at a corner.

Figure 4.7 Positions (a), headings (b), and curvatures (c) of the original w...

Figure 4.8 Data points of a track and its fitted result with polynomial spli...

Figure 4.9 Illustration of tracking errors for (a) discrete waypoint path de...

Chapter 5

Figure 5.1 Generated waypoints for a manually driven straight path.

Figure 5.2 Internal and external forces acting on elastic band.

Figure 5.3 Some path modification examples for different path and pedestrian...

Figure 5.4 Illustrations of selected NHTSA scenarios A (sudden appearance be...

Figure 5.5 Picture of the experimental vehicle at testing grounds.

Figure 5.6 Illustration for test scenario A.

Figure 5.7 HIL and experimental vehicle paths for scenario A testing.

Figure 5.8 Tracking errors for scenario A testing.

Figure 5.9 Illustration for test scenario B.

Figure 5.10 HIL and experimental vehicle paths for scenario B testing.

Figure 5.11 Tracking errors for scenario B testing.

Figure 5.12 Illustration for test scenario C.

Figure 5.13 HIL and experimental vehicle paths for scenario C testing.

Figure 5.14 Tracking errors for scenario C testing.

Figure 5.15 Interpolation conditions of the endpoints.

Figure 5.16 (a) Optimal polynomial curve, (b) curvature

κ

(

λ

), and ...

Figure 5.17 Table‐lookup path planning procedure.

Figure 5.18 Comparison of table‐lookup approach and optimization solution: (...

Figure 5.19 ROS visualization of LIDAR data with surrounding environment....

Figure 5.20 Preview path collides with obstacles.

Figure 5.21 Illustration of collision‐free target points generation.

Figure 5.22 Original and updated path segments along with their orders.

Figure 5.23 Online path planning result with an obstacle at a road curve....

Figure 5.24 A zoom‐in of Figure 5.23 showing the widest potential collision ...

Figure 5.25 Curvature of the original and updated path and their correspondi...

Figure 5.26 (a) Steering control, (b) lateral and look‐ahead error, (c) head...

Figure 5.27 Parameterization of the curvature control with a spline versus a...

Figure 5.28 Distribution of error coefficient

ε

k

in favor of early curv...

Figure 5.29 Optimized results in the spline or polynomial representation, re...

Figure 5.30 Different driving styles by adjusting relative weights in the ut...

Figure 5.31 Visualization of the lookup table generation process at one init...

Figure 5.32 SQP iterations until convergence from the initial guess (a) befo...

Figure 5.33 Sampling of jerk‐optimal quintic trajectories to reach the targe...

Figure 5.34 Trajectory representation in Cartesian and Frenet frame.

Figure 5.35 Trajectory candidates with optimized curvature controls in the F...

Figure 5.36 Application of the separating axis theorem to (a) collision chec...

Figure 5.37 Constrained optimization results of trajectory and curvature con...

Figure 5.38 Continuously planned and vehicle actual trajectories at icy road...

Figure 5.39 Estimated road friction coefficient fed to the trajectory planni...

Figure 5.40 Stroboscopic views of the selected trajectory among the dynamica...

Figure 5.41 Comparison of vehicle states of speed, heading, steering wheel a...

Figure 5.42 Two road segments selected for evaluation: a curved road with 25...

Figure 5.43 Collision avoidance under velocity keeping mode.

Figure 5.44 Wait for lane change due to roadblock ahead. Source: Zhu and Aks...

Figure 5.45 Merging into traffic.

Chapter 6

Figure 6.1 Path‐tracking plant with steering input and disturbances.

Figure 6.2 DOB curvature rejection filter loop for path‐tracking of an auton...

Figure 6.3 Wind force causing yaw moment disturbance for the vehicle.

Figure 6.4

Q

filter frequency responses.

Figure 6.5 Simulation results for yaw moment disturbance rejection case stud...

Figure 6.6 Simulation results for road curvature rejection case study, first...

Figure 6.7 Simulation results for road curvature rejection case study, secon...

Figure 6.8 Simulation results for road curvature rejection case study, third...

Figure 6.9 Loop structure for road curvature rejection with DOB in a control...

Figure 6.10 Equivalent form of DOB compensated path‐tracking loop under feed...

Figure 6.11 Simulation results for road curvature rejection with DOB in a co...

Figure 6.12 Simulation results for road curvature rejection with DOB in a co...

Figure 6.13 Simulation results for road curvature rejection with DOB in a co...

Figure 6.14 Loop structure for model regulation with DOB case study.

Figure 6.15 Frequency response analysis for model regulation with DOB case s...

Figure 6.16 Disturbance rejection properties comparison for disturbances wit...

Figure 6.17 Different

Q

filter frequency responses.

Figure 6.18 DOB curvature rejection filter with steering actuator time delay...

Figure 6.19 Illustration of stability robustness of DOB against time delay....

Figure 6.20 Simulation results for road curvature rejection with DOB with ti...

Figure 6.21 Simulation results for road curvature rejection with DOB with ti...

Chapter 7

Figure 7.1 Preview references in the moving‐vehicle framework.

Figure 7.2 Data of yaw rate measurement and Kalman filter estimation results...

Figure 7.3 MPC framework with Kalman filter for path following.

Figure 7.4 Vehicle experiment results showing the waypoints adjustment to av...

Figure 7.5 Measurement of vehicle speed, steering wheel angle, and yaw rate ...

Figure 7.6 Vehicle experiment results showing real‐time waypoints adjustment...

Figure 7.7 Measurement of vehicle speed, steering wheel angle, and yaw rate ...

Figure 7.8 Region

defined in the

s

‐plane.

Figure 7.9 Pole region

in the

s

‐plane with specifications. Source: Zhu et ...

Figure 7.10 An exemplary region

for transfer function

G

(

s

).

Figure 7.11 Illustration of critical condition of mixed sensitivity constrai...

Figure 7.12 Parameter‐space based robust gain‐scheduling design diagram. Sou...

Figure 7.13 Single‐track car model and its deviation from the desired path....

Figure 7.14 Tire lateral forces

F

y

and tire saturation parameter

η

with...

Figure 7.15 Uncertainty region of plant parameters and selected representati...

Figure 7.16 System diagram with the controller and the plant. Source: Zhu et...

Figure 7.17 Mappings of boundaries

, mixed sensitivity point conditions, an...

Figure 7.18 Feasible regions and selected control parameters (

l

s

,

k

p

) at eac...

Figure 7.19 Gain scheduling control law with control parameters (

l

s

,

k

p

) for...

Figure 7.20 Root set for the designed gain‐scheduling control. Source: Zhu e...

Figure 7.21 Additive parametric uncertainty at

v

x

...

Figure 7.22 Mixed sensitivity constraint validated at different speeds. Sour...

Figure 7.23 Hardware‐in‐the‐loop equipment setup.

Figure 7.24 Tracking trajectories with different vehicle masses at 20 m/s sp...

Figure 7.25 Tracking trajectories at different speeds with vehicle mass 1850...

Figure 7.26 Trajectory comparison of gain‐scheduling versus fixed gain contr...

Figure 7.27 Vehicle sideslip and steering of gain‐scheduling and fixed gain ...

Figure 7.28 Experimental vehicle on the test road. Source: Zhu et al. [4] (©...

Figure 7.29 Longitudinal speed and acceleration of road test and HIL simulat...

Figure 7.30 Vehicle trajectory at the slalom path (“PS” – Parameter‐Space de...

Figure 7.31 Phase portraits of vehicle sideslip angle (

β

,

) as well as...

Figure 7.32 Distance to the boundary line.

Figure 7.33 Phase portraits of vehicle sideslip angle and its rate (

β

,

Figure 7.34 Braking force distribution with respect to the direction of fron...

Figure 7.35 Path tracking results with front steering alone versus combined ...

Figure 7.36 Comparison of vehicle yaw rate (a) and lateral acceleration (b) ...

Figure 7.37 Comparison of phase portraits at speed 20 m/s and

μ

 = 0.5. ...

Figure 7.38 Comparison of request extra yaw moment from yaw‐rate and sidesli...

Figure 7.39 Wheel cylinder pressures for individual brake for the (a) yaw‐ra...

Guide

Cover Page

Table of Contents

Autonomous Road Vehicle Path Planning and Tracking Control

Title Page

Copyright

Dedication

About the Authors

Preface

List of Abbreviations

Begin Reading

Index

WILEY END USER LICENSE AGREEMENT

Pages

ii

iii

iv

v

xi

xii

xiii

xv

xvii

xviii

xix

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

221

222

223

225

226

227

228

229

230

231

i

232

IEEE Press445 Hoes LanePiscataway, NJ 08854

IEEE Press Editorial BoardEkram Hossain, Editor in Chief

Jón Atli Benediktsson

Anjan Bose

David Alan Grier

Elya B. Joffe

Xiaoou Li

Lian Yong

Andreas Molisch

Saeid Nahavandi

Jeffrey Reed

Diomidis Spinellis

Sarah Spurgeon

Ahmet Murat Tekalp

Autonomous Road Vehicle Path Planning and Tracking Control

Levent GüvençBilin Aksun‐GüvençSheng ZhuŞükrü Yaren Gelbal

Copyright © 2022 by The Institute of Electrical and Electronics Engineers, Inc. All rights reserved.

Published by John Wiley & Sons, Inc., Hoboken, New Jersey.

Published simultaneously in Canada.

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per‐copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750‐8400, fax (978) 750‐4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748‐6011, fax (201) 748‐6008, or online at http://www.wiley.com/go/permission.

Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762‐2974, outside the United States at (317) 572‐3993 or fax (317) 572‐4002.

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.

Library of Congress Cataloging‐in‐Publication Data:

Names: Güvenç, Levent, author. | Aksun‐Güvenç, Bilin, author. | Zhu,  Sheng author. | Gelbal, Şükrü Yaren, author. |  John Wiley & Sons, publisher.

Title: Autonomous road vehicle path planning and tracking control / Levent  Güvenç, Bilin Aksun‐Güvenç, Sheng Zhu, Şükrü Yaren Gelbal.

Description: Hoboken, New Jersey : Wiley‐IEEE Press, [2022] | Includes  bibliographical references and index.

Identifiers: LCCN 2021047513 (print) | LCCN 2021047514 (ebook) | ISBN  9781119747949 (cloth) | ISBN 9781119747956 (adobe pdf) | ISBN  9781119747963 (epub)

Subjects: LCSH: Automated vehicles–Design and construction. | Automated vehicles–Collision avoidance systems. | Mathematical optimization–Industrial applications.

Classification: LCC TL152.8 .G88 2022 (print) | LCC TL152.8 (ebook) | DDC 629.04/6–dc23/eng/20211101

LC record available at https://lccn.loc.gov/2021047513

LC ebook record available at https://lccn.loc.gov/2021047514

Cover Design: Wiley

Cover Images: Background image © Pobytov/Getty Images,

© Gorodenkoff/Shutterstock

Levent Güvenç and Bilin Aksun‐Güvenç dedicate this book to their parents and to their son Kunter Güvenç.

Sheng Zhu dedicates this book to his beloved family.

Şükrü Yaren Gelbal dedicates this book to his beloved family.

About the Authors

Levent Güvenç is a professor of mechanical and aerospace engineering at the Ohio State University (OSU) with a joint appointment at the electrical and computer engineering department. He is a member of the International Federation of Automatic Control (IFAC) Technical Committees on Automotive Control; Mechatronics; and Intelligent Autonomous Vehicles and the IEEE Technical Committees on Automotive Control (past member); and Intelligent Vehicular Systems and Control (founding member). During 1996–2011, Levent worked in the mechanical engineering department of Istanbul Technical University where he was the director of the European Union Framework Research Programme 6 funded Center of Excellence on Automotive Control and Mechatronics. He served as department chair of mechanical engineering at Istanbul Okan University from 2011 to 2014, was the founder and director of Mekar Mechatronics Research Labs and coordinator of team Mekar in the 2011 Grand Cooperative Driving Challenge. He is the co‐founder and co‐director of the Automated Driving Lab at the Ohio State University. He is the co‐author of 1 edited volume, 3 books including this one, 4 book chapters, and more than 220 publications in major journals and conference proceedings. He is an ASME fellow. He was the chair of the 2007 IEEE Intelligent Vehicles Symposium and has been a regular program committee member of IEEE Intelligent Vehicles and Intelligent Transportation Systems conferences. Levent has been working on ADAS and connected and automated driving system development with the automotive industry for more than two decades. He is the co‐inventor of six patents including two European patents, developed jointly with the automotive industry. He also has two provisional patent applications at OSU. His work on connected and autonomous driving has resulted in nine research prototype vehicles (two of them with a major automotive OEM) with different levels of autonomy, the last four being in the Automated Driving Lab at OSU. Levent was a member of the USDOT Smart Columbus project workgroups on Autonomous Electric Vehicles and the Connected Vehicle Environment.

Bilin Aksun‐Güvenç is a Research Professor in the Department of Mechanical and Aerospace Engineering of the Ohio State University since September 2017. She joined the Ohio State University in September 2014 as a Visiting Professor in the Department of Mechanical and Aerospace Engineering after working in Istanbul Technical University (17 years) and Istanbul Okan University (3 years) as a professor. Her expertise is in automotive control systems – primarily autonomous vehicles, connected autonomous vehicles, adaptive cruise control, cooperative adaptive cruise control, collision avoidance systems, cooperative collision avoidance systems, electronic stability control, vehicle dynamics controllers, intelligent transportation systems, smart mobility, and smart cities. She has a long history of working as a principal investigator and researcher in several automotive industry projects for major automotive OEMs. She is the co‐inventor of two European patents on yaw dynamics stability control and yaw rate virtual sensing for road vehicles, respectively. She is also the co‐inventor of two provisional patent applications on autonomous vehicle safety testing and autonomous driving function evaluation and development. She is a member of the IFAC technical committees on Automotive Control (since 2005) and on Mechatronics (since 2008). She is the author of 2 books including this one, 2 book chapters, and over 130 publications in technical journals and conferences. Prof. Bilin Aksun‐Güvenç participated in the NIST Global Cities Technical Challenge technical clusters Smart Shuttle and SMOOTH with the City of Columbus. Prof. Aksun‐Güvenç led the Connected Electric Vehicle Deployment project which was part of the cost share contribution of the Ohio State University to the Smart Columbus project. The project developed pre‐deployment MIL and HIL simulation evaluation of autonomous shuttle operation in the Linden Residential Area before the actual deployment was allowed to start. She is the co‐founder and co‐director of the Automated Driving Lab at the Ohio State University.

Sheng Zhu received the BS and MS degrees in automotive engineering from Beijing Institute of Technology and Tongji University, China, in 2012 and 2015, respectively. He received his PhD in 2020 in mechanical engineering from the Ohio State University, Ohio, USA. His research interests include on‐road trajectory planning and robust control with HIL and experiment applications in automated vehicles under all‐weather conditions, with over 10 research articles published and presented in peer‐reviewed journal and conferences. Dr. Zhu's honors include recipient of National Scholarship of China in 2009, and graduation with the highest GPA in automotive engineering, Beijing Institute of Technology, 2012. Dr. Zhu currently works as a software engineer on planning and control at DeepRoute.ai in developing level‐4 autonomous driving technology, with project development experience including four‐wheel‐driven vehicle control, autonomous valet parking, low‐level torque steer control, online calibration system, weight estimation and compensation, etc.

Şükrü Yaren Gelbal is a PhD student in the Ohio State University Electrical and Computer Engineering Department and a Graduate Research Associate in the Automated Driving Lab. Before joining the Ohio State University as a visiting scholar then pursuing his PhD, he received the BS degree in mechatronics engineering from Okan University, Istanbul, Turkey in 2014 and MS degree in mechatronics engineering from Istanbul Technical University, Istanbul, Turkey in 2017. During his studies, he worked in numerous industry‐funded, government‐funded, and non‐funded research projects while authoring and co‐authoring more than 30 papers with overall more than 150 citations. From 2013 to 2014, he was a research intern with the Automotive Controls and Mechatronics Research Lab in Okan University. From 2015 to 2016, he was a researcher at Istanbul Technical University for a project which was financially supported by the Scientific and Technological Research Council of Turkey. His research experience includes software development, software and hardware implementation for CAVs and UAVs, HIL simulations, collision avoidance, V2X communication, and VRU safety. Mr. Gelbal's honors include researcher scholarship by Scientific and Technological Projects Funding Program supported by the Scientific and Technological Research Council of Turkey for Master of Science students, 2015, graduation with High Honors from mechatronics engineering, Okan University, 2014, graduation with the Highest GPA from mechatronics engineering, Okan University, 2014.

Preface

Autonomous driving of road vehicles is an important topic with an increasing number of successful real‐world applications and the expected introduction of series production vehicles with Society of Automotive Engineers (SAE) Level L3 and SAE Level L4 autonomous driving capabilities by automotive companies around the world into the market. This trend has considerably increased both academic and industry research in this area.

The most fundamental task of an autonomous vehicle is the ability to plan and follow a path while avoiding collisions. Uncertainties in the environmental conditions, vehicle dynamics, vehicle load, and load distribution and the range of required speed from very low speeds to highway driving speeds require the path tracking and collision mitigation controls to be robust. This book contributes to this area by presenting the recent research results of the authors in path planning and robust path‐tracking control. The methods presented in the book are applicable in real life, having been tested in a realistic hardware‐in‐the‐loop simulation environment and in road testing with a research‐level autonomous vehicle along with the usual model‐in‐the‐loop simulations.

The target audience for this book is both researchers and practitioners working on autonomous vehicle motion planning and control, with the main target community being control scientists and engineers working on autonomous driving. This book focuses on the applications of robust motion control to the autonomous driving part of the automotive area. Students, especially graduate students, in this research area will also find the book of interest to them. Chapters 1, 2, 4, and 5 are based partially on material taught in courses by Prof. Levent Güvenç and Prof. Bilin Aksun‐Güvenç. These include the Ohio State University courses ECE 5553 Autonomy in Vehicles (graduate and senior undergraduate) and ME 8322 Vehicle System Dynamics and Control (graduate) and the past Istanbul Technical University and Istanbul Okan University courses on Automotive Control Systems (graduate) and Vehicle Dynamics and Control (undergraduate). The doctoral research work of Drs. Sheng Zhu and of Şükrü Yaren Gelbal (in progress) at the Ohio State university have also contributed significantly to this book.

The authors' graduate students and research collaborators have also made contributions to several topics covered in the book through joint publications and their names may be found in the references cited at the end of each chapter.

We hope that this book will be a useful reference for students, researchers, and practitioners interested in or working in the autonomous vehicle path planning, path‐tracking, and collision avoidance areas.

Columbus, OH, USA2021

Levent GüvençBilin Aksun‐GüvençSheng ZhuŞükrü Yaren Gelbal

List of Abbreviations

ABS

anti‐lock brakes

ACC

adaptive cruise control

ADAS

advanced driver assistance systems

AEB

automatic emergency braking

API

application programming interface

AV

autonomous vehicle

BFGS

Broyden–Fletcher–Goldfarb–Shanno

BNP

Bakker–Nyborg–Pacejka

BSM

basic safety message

CALTRANS

California Department of Transportation

CAN

controller area network

CC

cruise control

CDOB

communication disturbance observer

CMU

Carnegie Mellon University

CNN

convolutional neural network

CPU

central processing unit

CRB

complex root boundary

CV

connected vehicle

DDOB

double disturbance observer

DLC

double‐lane change

DOB

disturbance observer

DOF

degrees of freedom

DP

dynamic programming

DYC

direct yaw‐moment control

ESC

electronic stability control

ESP

electronic stability program

FRM

frequency response magnitude

FSM

finite state machine

FWS

front wheel steering

GCDC

grand cooperative driving challenge

GPS

global positioning system

GPU

graphical processing unit

HAD

highly automated driving

HIL

hardware‐in‐the‐loop

IMU

inertial measurement unit

IRB

infinite root boundary

ITS

intelligent transportation systems

JOSM

Java OpenStreetMap

KKT

Karush–Kuhn–Tucker

LCAS

lane change assistance system

LIDAR

laser imaging, detection, and ranging

LKAS

lane keeping assistance system

LKS

lane keeping system

LMI

linear‐matrix‐inequality

LPV

linear parameter‐varying

MABx

dSPACE MicroAutoBox

MF

magic formula

MIL

model‐in‐the‐loop

MPC

model predictive control

MSTE

mean squared tracking error

NAHSC

National Automated Highway System Consortium

NHTSA

National Highway Traffic Safety Administration

NLOS

non‐line‐of‐sight

NPC

non‐player character

OBU

on‐board unit

ORU

other road users

OS

overshoot

OSM

OpenStreetMap

P

proportional controller

PATH

California partners for advanced transit and highways

PCL

point cloud library

PD

proportional‐derivative controller

QP

quadratic programming

RALPH

rapidly adapting lateral position handler

RCA

Radio Corporation of America

RGB

red‐green‐blue

RLS

recursive least squares

RMS

root mean square

ROS

robot operating system

RRB

real root boundary

RRT

rapidly exploring random‐trees

RSU

roadside unit

RTK

real‐time kinematics

RWS

rear wheel steering

SAE

Society of Automotive Engineers

SCMS

security credential management system

SIL

software‐in‐the‐loop

SISO

single‐input single‐output

SLAM

simultaneous localization and mapping

SOTIF

safety of the intended functionality

SQP

sequential quadratic programming

SRI

Artificial Intelligence Center of Stanford Research Institute

SVM

support vector machine

TCS

traction control system

V2X

vehicle‐to‐everything

VRU

vulnerable road user

VSC

vehicle stability control

VVE

vehicle‐in‐virtual‐environment

XIL

X‐in‐the‐loop

1Introduction

This chapter provides an introduction to the whole book. After a section on motivation and introduction, a brief history of automated driving is presented, followed by how Advanced Driver Assistance Systems (ADAS) naturally evolved into autonomous driving functions. Some past and current autonomous driving architectures are presented using examples from the field. A literature review section where the key papers and more recent developments in path planning and robust path‐tracking control for autonomous road vehicles, also including the relevant literature on cybersecurity, and how it relates to autonomous vehicle path planning and tracking, are summarized next. This is followed by a section on the scope of the book, briefly detailing what is covered in each chapter. The chapter ends with a brief summary and concluding remarks.

1.1 Motivation and Introduction

The race toward series produced autonomous road vehicles has been rapidly progressing during the last decade. Most automotive OEMs and technology companies had promised or forecasted autonomous driving models by the year 2020, two years before the publication date of this book. This obviously did not take place. While we do not have truly autonomous driving vehicles that the public can currently buy, the currently available lane keeping, adaptive cruise control (ACC), emergency braking systems, traffic jam assistants, and their extended versions in some vehicles allow an almost autonomous highway driving experience under ideal conditions [1]. Autonomous shuttle service has been successfully deployed in a lot of different geofenced areas worldwide [2–4]. Large‐scale autonomous taxi service is about to start in several countries in Asia soon, using drive‐by‐wire vehicles retrofitted with sensors and control systems [5]. Autonomous vehicle races have also been increasing around the world [6]. Autonomous delivery vehicles and autonomous truck platoons are also technologies with many successful, limited‐scale deployments [7,8]. Automotive OEMs were planning to introduce autonomous products for the fleet market first, before making them available to the general public. Introduction of autonomous vehicle fleets that can also be used as ride hailed taxis is now expected by the year 2023 even though there may still be delays considering the failed predictions of the recent past. The current technology of traditional and nontraditional automotive OEMs and technology companies like Google's Waymo, and similar ones is sufficiently advanced for nearly full driverless operation in well‐mapped environments under ideal conditions. The relatively smaller percentage of nonideal conditions and uncertain environments make it difficult to implement full‐scale autonomous driving under arbitrary conditions and environments.

Figure 1.1 Categories of autonomous driving according to SAE.

Autonomous road vehicles have been categorized into six categories by the Society of Automotive Engineers (SAE) as shown in Figure 1.1[9]. Currently available automated driving technology in series produced vehicles falls under Level 2 which is partial automation. Level 2 partial automation is achieved in series production vehicles with lane‐centering control for steering automation and ACC and collision avoidance for automation in the longitudinal direction. L3 partial automation is characterized by all driving actuators being automated and the presence of a driver who can intervene when necessary. Recently introduced autopilot systems for cars are examples of conditional automation where the car takes care of driving in some driving modes like highway driving but the human operator is always in the driver seat to take over control if necessary. The Highway Chauffeur is a Level 3 autonomous highway driving system in which almost all highway driving functions are carried out autonomously, but the driver is needed to take over if something goes wrong or might go wrong like a lane change maneuver [10]. The Highway Chauffeur is currently available technology for series produced vehicles and uses an eHorizon electronic map to take care of driving on the highway until the chosen exit is reached. The Highway Pilot is a Level 4 autonomous driving extension of the Highway Chauffeur [11]. The driver is still in the driver's seat but the vehicle can perform highway driving completely autonomously without the need for driver interaction. Highway Pilots are expected to enter the market after 2022 [12].

In Level 5 driving automation, there is no need for a driver as the vehicle takes care of all driving tasks autonomously. It is clear that SAE Level 4 and Level 5 autonomous vehicles have to be capable of making their own decisions based on situational awareness using perception sensors and decision‐making algorithms to satisfy the fixed mission of following the highway between initial and final destination locations. This includes planning their route once the destination point is specified and taking care of path planning, path‐tracking control, and collision avoidance maneuvering, if needed, autonomously. This same approach is also needed for the lower speed autonomous driving in urban city environments which is a much more complicated situation due to the many other actors like vulnerable road users being present and more unexpected situations being likely to occur. This book treats path planning, path tracking control, and collision avoidance maneuvering for both urban and highway autonomous driving and also treats pedestrian collision avoidance of autonomous driving in the context of the urban application.

Automated driving shuttles in smart cities that are used for solving the first‐mile and last‐mile problem are other well‐known, emerging examples of autonomous road vehicles [11]. These shuttles operate at relatively lower speeds which definitely improves safety levels while also creating a traffic bottleneck around them. In comparison to limited access highway operation, these shuttles operate in significantly less‐structured environments with unpredictable interaction with vulnerable road users such as pedestrians, bicyclists, and scooters. The roads they use involve pedestrian crosswalks, intersections with or without traffic lights, roundabouts, and sharper turns as lower speed of operation is possible. Successful applications of these low‐speed autonomous shuttles exist in fixed routes. The whole route needs to be mapped in advance and extra landmarks in the form of signage have to be added in some cases as scan matching of the recorded map is used for localization of these autonomous shuttles. Level 4 like autonomous driving of these shuttles is achieved during the segments of the route without intersections and unexpected interactions with other road users. The safety driver takes over control of the vehicle in intersections and during unexpected events. This is called assisted autonomy and is currently necessary for safe operation. True Level 4 autonomous driving capability of these low‐speed urban environment autonomous vehicles is expected to be realized in the near future.

The most fundamental task of an autonomous vehicle is the ability to plan and follow a path while avoiding collisions. Path planning is optimized to make sure that the resulting trajectories have comfortable motion with limited acceleration and jerk. Uncertainties in environmental conditions, vehicle dynamics, vehicle load, and load distribution and the range of required speed from very low speeds for urban driving to highway driving speeds require the path tracking and collision mitigation controls to be robust. The motivation of this book is to contribute to this very important area of autonomous driving by presenting recent research results in path planning and robust path‐tracking control. Robustness is achieved through two different approaches. The first one is regulation of the path following dynamic model to reject the uncertainties and disturbances and to handle the variable time delays that are present. The second approach is to use a robust feedforward and feedback controller combination to achieve guaranteed performance. The presence of static or moving obstacles such as other cars, pedestrians, and bicyclists is also treated by presenting methods for modifying the path to avoid such collisions in realistic applications. The methods presented in the book are applicable in real life, having been tested in a realistic hardware‐in‐the‐loop simulation environment and in road testing with a research‐level autonomous vehicle in addition to the usual model‐in‐the‐loop simulations.

1.2 History of Automated Driving

Contrary to popular belief, the origins of autonomous driving and automated vehicles go back all the way to the 1920s. Radio‐controlled cars were the novelty in the 1920s while 1960s and 1970s saw the emergence of cable‐controlled cars, actually and unknowingly taking a step backwards. Computer‐controlled cars resembling today's autonomous vehicles started emerging in a very rough form in the 1980s and 1990s. In the first driverless car experiments of the 1920s, an antenna was mounted on the car which was driven by an external operator using radio signals, much like radio‐controlled toys. It should be noted that this remote operation forms the basis of some current driverless vehicles being followed by a second vehicle whose operators can take over control and intervene, if necessary. The presence of a safety operator, whether in a nearby other vehicle or in the driver's seat, is one of the current major limitations of autonomous driving. The car in the 1920s example obviously had to be in the field of vision of the external operator and also had to be operated at low speeds. A totally different approach was taken starting in the 1960s to get rid of the safety driver. Cables carrying electricity and generating a magnetic field were embedded in the roads. A downward‐looking magnetic pickup in the car was used to follow this magnetic field much like the way toy cars follow reflective tape fixed on the ground. This approach obviously did not work in the long run due to the tremendous work and cost required for the necessary road infrastructure. The first computer‐controlled cars that were developed later used camera and later also radar to do lane keeping and speed alterations much like today's lane‐centering and ACC systems to automatically follow roads and cars. The computational power available was very limited as compared to what is available now with our robust operating systems and our fast CPUs and GPUs with large memory and storage capability.

The first documented radio controlled car was produced in 1925 by Houdina Radio Control, a radio equipment firm. The car called the Linrrican Wonder traveled through a traffic jam in New York City and was controlled by a transmitting antenna. The car behind this “phantom auto” sent signals to the antenna, and the signals worked on small electric motors that actuated the necessary pedals and steering. The Linrrican Wonder was obviously also a drive‐by‐wire vehicle. It was driverless in that the driver was in the following vehicle. The actuator positions of this following vehicle were mimicked by the driverless vehicle in a master–slave configuration. This is similar to the master–slave configuration that is used in some platooning applications with the difference that the master is the following vehicle and not the lead one which will obviously introduce unnecessary phase lag to the coupled system due to the communication time delay involved and this delay will destabilize the system at higher speeds. Nevertheless, this radio‐controlled phantom lead car followed by the second control signal‐generating car is an early example of automated platooning. The control signals rather than vehicle acceleration are sent to the next vehicle as is proposed in the recent connected cruise control (CC) concept. The human operator in the following car also acts like a cloud computer that computes and relays the necessary control commands to the driverless lead car. This radio‐controlled driverless car is illustrated in Figure 1.2. Note that a similar procedure of having a second car with operators follow a truly driverless car with passengers was used recently for safety purposes [13]. This radio‐controlled driverless vehicle concept from the 1920s also has similarities to the currently used phantom driver concept which is used to remotely operate a truly driverless vehicle in emergency situations that the autonomous driving system cannot handle [14]. The remote operator has full access to the sensor data and surround view from the driverless car and can operate the car using his/her steering and pedal inputs like manual driving. The variable communication delays involved create possible stability problems that need to be handled just like those that occur in telemanipulation. These delays are expected to become much smaller in magnitude with 6G communication which is the driving point behind current collaborative perception and awareness systems.

Figure 1.2 Illustration of radio‐controlled driverless car of 1920s.

There have also been successful proof‐of‐concept type implementations of Intelligent Transportation Systems (ITSs) technology dating back to 1925. Charles Adler was named the man who invented Intelligent Traffic Control a century too early [15]. He embedded magnetic plates in the road at a point before the road led into a sharp curve. He also prepared a car with a speed governor that would be activated as the car drove over the magnetic plate such that the vehicle engine would be commanded to slow down to 24 km/h. This is a very early hardware implementation of Curve Speed Warning [16] and automatic speed reduction system where the whole system has been implemented as a hardware solution.

The General Motors Futurama exhibit at the 1939 World's Fair in New York is viewed by many as the first large‐scale ITS and highway automation concept and vision [17]. The vision of Superhighways was demonstrated in this exhibit where the visitors would sit and watch a scaled down replica of a future city complete with highways and guided cars. This was a hard 3D model version of the current game engine rendered environments with traffic cosimulation [3,18–21]. These guided cars were navigated automatically using radio control and could enter curves at speeds up to 50 mph (80 km/h). This vision of Superhighways was presented as what would happen in 1960. In compliance with this idea, there was a desire in the automotive industry to have cable buried in lane centerlines in these envisioned superhighways to create a magnetic field that could be followed automatically by self‐driving vehicles. This approach was not adopted because of higher initial cost and higher maintenance cost during repairs. This vision was one of self‐driving based on infrastructure which is exactly why it failed.

The current infrastructure‐based technology that can be used for following lane centerlines is the use of a front‐looking camera to detect and track the lanes in a lane‐keeping assistance system. The problem with this technology is that the lanes cannot be detected when weather conditions degrade the camera performance or when the road is covered with snow, making the lanes invisible. Since the interstate roads were built without voltage carrying cable buried inside, automotive OEMs lost interest in self‐driving as the required infrastructure was not there. Researchers and some states like California continued on with this Superhighway idea, which changed its name to Automated Highway Systems, until the 1990s, but while there were a lot of successful demonstrations and a large literature of research results, larger‐scale use of this technology never happened as automotive companies did not want to develop cars dependent on nonexisting and costly infrastructure and as there was also no demand from car buyers. Nevertheless, different ways of embedding passive signals to be followed inside roads continued until recent years in the form of equally spaced metallic pins, reflective tape, and more recently smart paint [22,23]. Out of these, smart paint uses nanoparticles embedded inside normal paint used in lane markings and can be detected with a simple sensor. It can also be detected if the road is covered with snow and offers a relatively cheap and highly accurate way of localization in campus like environments or geofenced urban areas for low‐speed autonomous driving applications as the preview distance is constrained to be small.

According to reference [24], the first self‐driving car was built and tested by the Radio Corporation of America (RCA) in 1957 on a 400 ft public highway in Lincoln, Nebraska. The steering wheel and pedals in this vehicle were replaced with a small joystick and an emergency brake while the vehicle speed and distance with the car in front were displayed on the dashboard. Self‐driving was activated by pressing the Electronic Drive button and the car would follow its lane and adjust its speed automatically. RCA was working on this technology around the 1950s. During the same time frame around 1956, GM shared its vision of a futuristic car driving autonomously in an automated highway using a professional video in the Motorama auto show. Automated highway driving involved a control center directing traffic much like today's aircraft traffic control centers.

According to Wetmore [25] GM and RCA developed a scaled version of an automated highway system by 1953. They used this scaled automated highway system to investigate how electronics can be used for path‐tracking and car following. Note that this is very similar to the scaled drive‐by‐wire vehicles with sensors like the F1TENTH vehicles [26,27] that a lot of academic research groups use for in‐the‐lab studies of autonomous driving and decision‐making. In 1958, GM built a prototype self‐driving vehicle with a pickup in front that would sense the alternating current of a wire embedded in the road [25]. This was used for localization with respect to the path to be followed and for generating the required corrective steering action. The research vehicle could take turns automatically without driver steering intervention. By 1960, GM had tested its research vehicles in test tracks with automatic path‐tracking, lane‐change maneuvering, and car following. Since the technology used was based on cable and signals buried under the road, these research efforts stopped after the interstates were built without this infrastructure required for automated highways.

The first university based self‐driving studies started around the 1960s just as the industry studies were ending. The same technology of a live wire either buried in the road or fixed on top of the road was used to steer a vehicle in a closed test area as shown in Figure 1.3 by Ohio State University researchers [28]. An analog computer was used to implement steering control and speed control as shown in Figure 1.4. This vehicle was converted into a drive‐by‐wire one and the analog joystick shown in Figure 1.5 was developed and used in the car for operator intervention. Turning the handle of the joystick left or right was used for steering in the same direction. Moving the joystick handle forward applied throttle while moving it backwards applied braking. In later years, longitudinal car following was implemented by actually connecting the two cars by wire and communicating the necessary information like distance and speed difference through this hard wire. Connected vehicles (CVs) actually meant being physically connected during 1960s. As compared to today's wireless networked car following systems, this wired and truly connected version obviously did not have the wireless communication delays that limit the minimum achievable time gaps of present platooning systems and cooperative ACC.

While the historical self‐driving car work presented until this point focused on automating the longitudinal and lateral motions of the on‐road vehicle, its controls, some very basic radar perception and road‐fixed cable based localization; work on autonomy or intelligence, i.e. decision‐making, was not being pursued owing mostly to the structured nature of highway driving. On the other hand, the mobile robotics community was using low‐speed mobile robots with simple mechanical designs to use camera sensors for perception of the environment and scene understanding to be used in decision‐making for autonomously navigating through previously unknown obstacles by planning and executing a collision free path. Unlike the highway driving application,