GNSS for Vehicle Control

无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com GNSS for Vehicle Control 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com For a list of titles in the Artech House GNSS Technology and Applications Series, please turn to the back of the book. 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com GNSS for Vehicle Control David M. Bevly Stewart Cobb a r techhouse . com 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the U.S. Library of Congress. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library. ISBN 13: 978-1-59693-301-9 Cover design by Vicki Kane © 2010 ARTECH HOUSE 685 Canton Street Norwood, MA 02062 All rights reserved. Printed and bound in the United States of America. No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in writing from the publisher. All terms mentioned in this book that are known to be trademarks or service marks have been appropriately capitalized. Artech House cannot attest to the ac- curacy of this information. Use of a term in this book should not be regarded as affecting the validity of any trademark or service mark. 10 9 8 7 6 5 4 3 2 1 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com � Contents Preface xiii Acknowledgments xvii 1 GNSS and Other Navigation Sensors 1 1.1 Global Navigation Satellite System (GNSS) 1 1.1.1 Description of a Typical GNSS 2 1.1.2 Simple (Pseudorange) GNSS Navigation 3 1.1.3 Differential GNSS Navigation 6 1.1.4 Precise (RTK) GNSS Navigation 8 1.1.5 Current and Future GNSS Constellations 11 1.2 Pseudolites 14 1.2.1 Pseudolite Basics 14 1.2.2 Pseudolite/GNSS Navigation 14 1.2.3 Differential Pseudolite/GNSS Navigation 15 1.2.4 Pseudolite Self-Synchronization 16 1.2.5 Stand-Alone Pseudolite Navigation 16 1.2.6 Conflicts with GNSS Frequencies 17 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com �i GNSS for Vehicle Control 1.3 Inertial Navigation Systems (INS) 18 1.3.1 Linear Inertial Instruments: Accelerometers 18 1.3.2 Angular Inertial Instruments: Gyroscopes 20 1.3.3 Ideal Inertial Navigation 21 1.3.4 Sensing Earth Effects 23 1.3.5 Inertial Instrument Errors 25 1.3.6 Inertial Error Propagation 30 1.4 Odometer Technology 31 1.4.1 Quantization 32 1.4.2 Wheel Slip 32 1.4.3 Wheel Radius Error 33 1.5 GNSS/Inertial Integration 34 References 35 2 Vision Aided Navigation Systems 39 2.1 Lane Positioning Methods 40 2.1.1 Lidar-Based Positioning 40 2.1.2 Camera-Based Positioning 42 2.2 Coordinate Frame Rotation and Translation 43 2.2.1 Two-Dimensional Rotations 44 2.2.2 Three-Dimensional Rotations 45 2.2.3 Coordinate Frame Translation 46 2.2.4 Global Coordinate Frame Rotations 47 2.3 Waypoint-Based Maps 48 2.4 Aiding Position, Speed, and Heading Navigation Filter with Vision Measurements 49 2.4.1 Two-Dimensional Map Construction 50 2.4.2 Measurement Structure 51 2.4.3 Checking Waypoint Map Position 51 2.4.4 Results 52 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com 2.5 Aiding Closely Coupled Navigation Filter with Vision Measurements 52 2.5.1 Three-Dimensional Map Construction 54 2.5.2 Measurement Structure 56 2.5.3 Checking Waypoint Map Position 58 2.5.4 Results 58 References 59 3 Vehicle Modeling 61 3.1 Introduction 61 3.2 SAE Vehicle Coordinates 61 3.3 Bicycle Model 63 3.3.1 Basics 63 3.3.2 Understeer Gradient 70 3.3.3 Four-Wheel Bicycle Model 71 3.4 Tires 74 3.4.1 Basics 74 3.4.2 Contact Patch and Slip 74 3.4.3 Tire Models 76 3.5 Roll Model 79 3.5.1 Free Body Diagram 79 3.5.2 Equation of Motion 80 3.5.3 State Space Representation 80 3.6 Additional Models Used in this Work 80 3.6.1 Two-Wheeled Vehicle 81 3.6.2 Trailer Model 82 3.7 Vehicle Model Validation 84 References 88 Contents �ii 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com �iii GNSS for Vehicle Control 4 Navigation Systems 91 4.1 Introduction 91 4.2 Kalman Filter 92 4.3 GPS/INS Integration Architectures 93 4.3.1 Loose Coupling 93 4.3.2 Close Coupling 94 4.4 Speed Estimation 95 4.4.1 Accelerometer and GPS 96 4.4.2 Accelerometer, GPS, and Wheel Speed 102 4.5 Heading Estimation 107 4.6 Position, Speed, and Heading Estimation 111 4.6.1 Coordinate Conversion 112 4.6.2 Accelerometer, Yaw Rate Gyroscope, GPS, and Wheel Speed 113 4.7 Navigation in the Presence of Sideslip 120 4.7.1 Generation of Sideslip 120 4.7.2 Sideslip Compensation with a Dual Antenna GPS Receiver 122 4.8 Closely Coupled Integration 130 References 143 5 Vehicle Dynamic Estimation Using GPS 145 5.1 Introduction 145 5.2 Sideslip Calculation 146 5.3 Vehicle Estimation 147 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com 5.4 Experimental Setup 148 5.4.1 Test Scenarios 148 5.5 Kinematic Estimator (Single GPS Antenna) 149 5.6 Kinematic Kalman Filter (Dual Antenna) 151 5.7 Tire Parameter Identification 154 5.8 Model-Based Kalman Filter 160 5.8.1 Linear Tire Model 161 5.8.2 Nonlinear Tire Model 164 5.8.3 Estimator Accuracies 170 5.9 Conclusions 171 Acknowledgments 172 References 172 6 GNSS Control of Ground Vehicles 175 6.1 Introduction 175 6.2 Vehicle Model 175 6.3 Speed Controller 179 6.4 Vehicle Steering Control 181 6.4.1 Classical Steer Angle Controller 181 6.4.2 Classical Yaw Rate Controller 182 6.5 Waypoint Control 185 6.5.1 Heading Model 185 6.5.2 Heading Error Calculations 186 6.5.3 Heading Control 187 6.5.4 Simulation Results 190 Contents ix 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com x GNSS for Vehicle Control 6.6 Lateral Control 192 6.6.1 Error Calculation 193 6.6.2 Lateral Position Model 198 6.6.3 Lateral Position Control 200 6.6.4 Simulation Results 203 6.7 Implement/Trailer Control 203 6.7.1 Trailer Model 204 6.7.2 Error Calculation 206 6.7.3 Trailer Control 208 6.7.4 Simulation Results 210 References 212 7 Pseudolites for Vehicle Navigation 215 7.1 Pseudolite Applications 215 7.1.1 Open-Pit Mining 216 7.1.2 Construction Sites 218 7.1.3 Urban Navigation 218 7.1.4 Indoor Applications 219 7.2 Pseudolite Systems 221 7.2.1 IntegriNautics IN400 221 7.2.2 Novariant Terralite XPS System 223 7.2.3 Locata LocataLites 225 References 226 Appendix Estimation Methods 229 A.1 Introduction 229 A.2 System Model 229 A.3 Discretization 231 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com A.4 Least Squares 233 A.5 Weighted Least Squares 236 A.6 Recursive Weighted Least Squares 243 A.7 Kalman Filter 246 A.8 Extended Kalman Filter 249 A.9 Initialization 252 References 252 About the Authors 253 Index 257 Contents xi 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com xiii Preface As Global Navigation Satellite Systems (GNSS) such as GPS have grown more pervasive, the use of GNSS to automatically control ground vehicles has drawn increasing interest. From autonomously driven vehicles such as those demonstrated in the DARPA “grand challenges” to automatically steered farm tractors, automated mining equipment, and military unmanned ground vehicles (UGVs), practical and potential applications of GNSS to ground vehicles abound. This text provides an introduction to the concepts necessary to understand and contribute to this field. It has been said that navigation is “knowing where you are,” guidance is “knowing where you’re going,” and control is “knowing how to get there.” For example, suppose you are sitting at home one hot afternoon and you decide to ride your bicycle to the store to get an ice cream cone. First, consider navigation. You know that you’re starting at home. Fur- thermore, you know the names of all the streets all the way to the ice cream store and the names of all the nearby streets as well. At any point during the trip, you can look up at a street sign and know exactly where you are. You will have no trouble with navigation. Second, consider guidance. You know where the store is, and you can think of many routes to get from your house to the store. Some routes would keep you on the paved streets, but it might be faster to take a shortcut across the park. It would be even faster to take a shortcut across the river, but the 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com xi� GNSS for Vehicle Control water is too deep; you can only cross at the bridge. You choose a feasible route in your mind which meets all the constraints you know of. If you have to leave your original route along the way, due to a parade or a street fair, you can choose a new route and still reach the store. Guidance will not be a problem. Finally, consider control. You wheel your bike out the gate, remember- ing when you learned to ride it. At first, it wasn’t easy to keep your balance; you had to learn to steer in the direction you were falling as soon as you started to fall. You also had to learn to steer briefly away from any turn you wanted to make, to change your balance so that you could lean into the turn. But you understand those algorithms now, and you can implement them fast enough to get where you want to go without falling over. Your control of the bicycle is adequate. You set off for the store. Perhaps this example seems overly complicated. We humans perform most of these tasks unconsciously, most of the time. If we want to program automated systems to perform them, however, we must first understand them in detail. The purpose of this book is to develop an understanding of the navigation and control tasks for the special case of ground vehicles. The result of a navigation algorithm is the same for every vehicle; your position is your position, whether you’re on a bike, a bus, or a boat. However, each vehicle can have a different set of navigation sensors, and its navigation algorithm must be able to use the data from those sensors to compute the best possible position. Chapters 1, 2, and 7 of this book describe navigation sensors applicable to ground vehicles. Chapter 4 describes navigation algo- rithms that combine the data from these sensors to provide the best available estimate of position and velocity. Control algorithms are different for every vehicle; buses, boats, and bi- cycles are steered in very different ways. Chapter 3 describes mathematical models for various categories of vehicles, with a particular emphasis on com- mon four-wheeled vehicles. These models are used in Chapters 5 and 6 to de- velop vehicle control and estimation algorithms for tasks such as autonomous steering and electronic stability control (ESC). The problem of guidance is not so easy to generalize. Each type of ve- hicle has different constraints on the routes it can take. In addition, a given vehicle’s mission imposes other goals and constraints not shared even by simi- lar vehicles on different missions. Guidance is therefore largely beyond the scope of this text. The authors expect this book to be a useful introduction, for a gradu- ate engineer or perhaps an advanced undergraduate, to the problems of 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com navigating and controlling ground vehicles automatically. Because Global Navigation Satellite Systems (GNSS) are now widely available, Chapter 1 discusses concepts and navigation algorithms related to GNSS and to ground-based navigation transmitters known as “pseudolites.” Chapter 1 also introduces inertial measurement instruments, which sense acceleration and rotation directly; compasses, which sense heading relative to the Earth’s magnetic field; and odometers, which measure distance by counting the ro- tations of a wheel. Higher-level navigation algorithms, which combine data from these sensors, are presented in Chapter 4. Chapter 2 describes the use of “machine vision” algorithms to detect a vehicle’s position relative to road features such as lane markers, as seen through vehicle-mounted cameras and laser scanners (Lidar). These are espe- cially useful for lateral navigation of a vehicle on a road for Lane Departure Warning (LDW) or lane-keeping assistance. These navigation algorithms combine locally generated measurements from a camera or Lidar with global measurements from GPS and a map database to form a robust measurement of position in the lane. Chapter 3 introduces models for vehicles, tires, suspensions, and trail- ers. These models describe the behavior of highway vehicles such as passenger cars and SUVs as well as off-road vehicles such as farm tractors and unmanned ground vehicles (UGVs). This chapter describes the lateral and longitudinal dynamics that arise from these models and compares model data with mea- sured data for a particular vehicle. Chapter 4 describes methods for creating and updating an estimate of the navigation state using various combinations of the sensor measurements discussed in Chapter 1. These methods are based on the concept of the Kal- man filter (a tutorial review of filtering and estimation techniques is provided as the Appendix to this book). Again, results from a simulation model are compared to data measured on a particular vehicle. Chapter 5 describes methods for estimating parameters specific to four- wheeled vehicles, which are critical for modeling and control of vehicles. These estimated parameters can then be incorporated into a mathematical model of that specific vehicle for online vehicle modeling and vehicle control. Chapter 6 develops and analyzes control algorithms tuned to the models de- veloped in the previous chapters. Finally, Chapter 7 gives a detailed description of “pseudolites,” ground-based transmitters of signals similar to GNSS signals, which are useful for navigating ground vehicles in restricted areas such as open-pit mines. Preface x� 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com x�ii Acknowledgments The authors would like to thank their editor, Mark Walsh, for his patience and guidance through the publication process. David M. Bevly would like to thank the research team of current and former students in the GPS and Vehicle Dynamics Laboratory at Auburn University, specifically those who contributed to the chapters in this book. Additionally, I give thanks to the Lord my God, and His son Jesus, through whom all things are possible (Phil. 4:13). Stewart Cobb would like to thank his colleagues at Stanford University, IntegriNautics, and Novariant for providing a stimulating, challenging, but always friendly work environment; and his daughter for her tolerance and understanding when Daddy was playing with his computer rather than with her. 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com � 1 GNSS and Other Navigation Sensors Stewart Cobb and Benjamin Clark Ground vehicles can navigate using signals from external navigation systems such as the Global Positioning System (GPS) (Figure 1.1), or by using sig- nals from internal devices such as a compass, an odometer, a gyroscope, an accelerometer, or a full-blown inertial navigation system (INS). In practice, the most reliable and accurate navigation is obtained by combining data from all available sources, including static databases such as a digitized map. This chapter will discuss many of these sources of navigation data, and methods for combining them. 1.1 Global Navigation Satellite System (GNSS) There are at least two Global Navigation Satellite Systems (GNSSs) currently in existence, and several more are proposed. These systems are all similar in concept, differing in small details of signal frequencies, signal structure, and orbit design. The concepts presented in this chapter should apply to any GNSS system, regardless of those differences. Where specific details must be cited, the United States’ GPS will be used as an example. GPS is the most thoroughly studied GNSS, and the most useful at present. 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com � GNSS for Vehicle Control 1.1.1 Description of a Typical GNSS A GNSS receiver navigates by precisely measuring the range between its an- tenna and a set of transmitter antennas at precisely known locations, and then performing a triangulation algorithm to determine its position. This is not quite as easy as it sounds, because the transmitters are aboard satellites moving rapidly through space, and the measurements must be made with nanosecond precision. The space segment of a GNSS consists of a group, or constellation, of satellites in orbits that circle the Earth about twice per day. In order to pro- vide adequate signal coverage to the whole earth, a constellation typically consists of 20 to 30 satellites in three to six different orbital planes. Some systems also include satellites in geosynchronous orbits. The satellites broadcast microwave signals toward the Earth. Each satel- lite is far enough from Earth that its signal covers most of a hemisphere. Each signal consists of a carrier wave at a frequency near 1.6 GHz, modulated by a stream of digital bits at a rate of about 1 million bits per second (1 Mbps). The digital bits are generated in a way that is actually systematic but which appears random, and are called a pseudorandom noise code or PRN code. Each satellite has its own specific PRN code. The PRN code is itself modulated by digital navigation data at a slow rate (typically 50 bits per second). Figure 1.1 GPS satellite constellation, approximately to scale. 无水印完整版本下载：www.n-ebook.com Doc uCo m P DF Tria l ww w.pd fwiz ard. com GNSS and Other Navigation Sensors � The frequency of each satellite’s signal and the bit rate of its PRN code are controlled by an extremely precise clock (an atomic clock) on board the satellite. The uncompensated drift rate of each satellite’s clock is typically a few nanoseconds per day. The satellite signal is designed so that a receiver which “hears” the signal can read the exact time of the satellite’s clock at the instant the signal was transmitted, with an error of a few nanoseconds. Each GNSS has a master control center, which constantly listens to the satellite signals through receivers in several different locations. It uses this information to compute the exact orbits and clock drift corrections for all the satellites, and transmits this information to each satellite in turn. This information is then broadcast by the satellite as part of the navigation data message. Each user receiver can interpret the navigation data to determine the precise time (according to the whole GNSS system, not just a particular satellite clock) that a signal was transmitted from a satellite, and the precise position of the satellite (within a meter or so) when it was transmitted. At a particular instant (chosen according to its own internal clock), the receiver takes a snapshot of the clock readings of each satellite it can “hear.” It then subtracts the reading of its own internal clock from the readings of the satellite clocks. The difference is the time taken by each signal to travel from its satellite to the receiver. Assuming that the signal traveled at the speed of ligh