Available:*
Library | Item Barcode | Call Number | Material Type | Item Category 1 | Status |
---|---|---|---|---|---|
Searching... | 30000010312138 | TJ211.415 R63 2012 | Open Access Book | Book | Searching... |
On Order
Summary
Summary
Focusing on autonomous robotic applications, this resource offers a practical treatment of short-range radar processing for reliable object detection at the ground level. It demonstrates probabilistic radar models and detection algorithms specifically for robotic land vehicles.
Author Notes
Martin Adams is a professor in the department of electrical engineering and a member of the Advanced Mining Technology Centre (AMTC) at the University of Chile. He received his D.Phil. in engineering science at the University of Oxford.
John Mullane is a research scientist at Projective Space Pte. Ltd., Singapore. He holds a Ph.D. in electrical and electronic engineering from Nanyang Technological University.
Ehi Jose is a research scientist at Projective Space Pte. Ltd., Singapore. He holds a Ph.D. in electrical and electronic engineering from Nanyang Technological University.
Ba-Ngu Vo is a professor at the School of Electrical, Electronic and Computer Engineering at the University of Western Australia. He received his Ph.D. in engineering from Curtin University.
Table of Contents
Preface | p. xiii |
Acknowledgments | p. xv |
Acronyms | p. xvii |
Nomenclature | p. xxi |
Chapter 1 Introduction | p. 1 |
1.1 Isn't Navigation and Mapping with Radar Solved? | p. 1 |
1.1.1 Applying Missile/Aircraft Guidance Technologies to Robotic Vehicles | p. 2 |
1.1.2 Placing Autonomous Navigation of Robotic Vehicles into Perspective | p. 11 |
1.2 Why Radar in Robotics? Motivation | p. 12 |
1.3 The Direction of Radar-Based Robotics Research | p. 19 |
1.3.1 Mining Applications | p. 19 |
1.3.2 Intelligent Transportation System Applications | p. 21 |
1.3.3 Land-Based SLAM Applications | p. 24 |
1.3.4 Coastal Marine Applications | p. 25 |
1.4 Structure of the Book | p. 27 |
Reference | p. 28 |
Part I Fundamentals of Radar and Robotic Navigation | p. 33 |
Chapter 2 A Brief Overview of Radar Fundamentals | p. 35 |
2.1 Introduction | p. 35 |
2.2 Radar Measurements | p. 36 |
2.3 The Radar Equation | p. 38 |
2.4 Radar Signal Attenuation | p. 40 |
2.5 Measurement Power Compression and Range Compensation | p. 43 |
2.5.1 Logarithmic Compression | p. 44 |
2.5.2 Range Compensation | p. 44 |
2.5.3 Logarithmic Compression and Range Compensation During Target Absence | p. 45 |
2.5.4 Logarithmic Compression and Range Compensation During Target Presence | p. 47 |
2.6 Radar-Range Measurement Techniques | p. 51 |
2.6.1 Time-of-Flight (TOF) Pulsed Radar | p. 51 |
2.6.2 Frequency Modulated, Continuous Wave (FMCW) Radar | p. 53 |
2.7 Sources of Uncertainty in Radar | p. 58 |
2.7.1 Sources of Uncertainty Common to All Radar Types | p. 59 |
2.8 Uncertainty Specific to TOF and FMCW Radar | p. 68 |
2.8.1 Uncertainty in TOF Radars | p. 68 |
2.8.2 Uncertainty in FMCW Radars | p. 70 |
2.9 Polar to Cartesian Data Transformation | p. 72 |
2.9.1 Nearest Neighbor Polar to Cartesian Data Conversion | p. 73 |
2.9.2 Weighted Polar to Cartesian Data Conversion | p. 73 |
2.10 Summary | p. 76 |
2.11 Bibliographical Remarks | p. 76 |
2.11.1 Extensions to the Radar Equation | p. 76 |
2.11.2 Signal Propagation/Attenuation | p. 77 |
2.11.3 Range Measurement Methods | p. 78 |
2.11.4 Uncertainty in Radar | p. 78 |
References | p. 79 |
Chapter 3 An Introduction to Detection Theory | p. 81 |
3.1 Introduction | p. 81 |
3.2 Concepts of Detection Theory | p. 82 |
3.3 Different Approaches to Target Detection | p. 84 |
3.3.1 Non-adaptive Detection | p. 84 |
3.3.2 Hypothesis Free Modeling | p. 85 |
3.3.3 Stochastic-Based Adaptive Detection | p. 86 |
3.4 Detection Theory with Known Noise Statistics | p. 87 |
3.4.1 Constant P CFAR fa with Known Noise Statistics | p. 87 |
3.4.2 Probability of Detection P CFAR n with Known Noise Statistics | p. 88 |
3.4.3 Probabilities of Missed Detection P CFAR MD and Noise P CFAR n with Known Noise Statistics | p. 89 |
3.5 Detection with Unknown Noise Statistics-Adaptive CFAR Processors | p. 89 |
3.5.1 Cell Averaging-CA-CFAR Processors | p. 90 |
3.5.2 Ordered Statistics-OS-CFAR Processors | p. 94 |
3.6 Summary | p. 100 |
3.7 Bibliographical Remarks | p. 101 |
References | p. 102 |
Chapter 4 Robotic Navigation and Mapping | p. 105 |
4.1 Introduction | p. 105 |
4.2 General Bayesian SLAM-The Joint Problem | p. 107 |
4.2.1 Vehicle State Representation | p. 109 |
4.2.2 Map Representation | p. 112 |
4.3 Solving Robot Mapping and Localization Individually | p. 115 |
4.3.1 Probabilistic Robotic Mapping | p. 116 |
4.3.2 Probabilistic Robotic Localization | p. 116 |
4.4 Popular Robotic Mapping Solutions | p. 117 |
4.4.1 Grid-Based Robotic Mapping (GBRM) | p. 117 |
4.4.2 Feature-Based Robotic Mapping (FBRM) | p. 118 |
4.5 Relating Sensor Measurements to Robotic Mapping and SLAM | p. 120 |
4.5.1 Relating the Spatial Measurement Interpretation to the Mapping/SLAM State | p. 121 |
4.5.2 Relating the Detection Measurement Interpretation to the Mapping/SLAM State | p. 122 |
4.6 Popular FB-SLAM Solutions | p. 124 |
4.6.1 Bayesian FB-SLAM-Approximate Gaussian Solutions | p. 124 |
4.6.2 Feature Association | p. 126 |
4.6.3 Bayesian FB-SLAM-Approximate Particle Solutions | p. 128 |
4.6.4 A Factorized Solution to SLAM (FastSLAM) | p. 129 |
4.6.5 Multi-Hypothesis (MH) FastSLAM | p. 130 |
4.6.6 General Comments on Vector-Based FB SLAM | p. 130 |
4.7 FBRM and SLAM with Random Finite Sets | p. 133 |
4.7.1 Motivation: Why Random Finite Sets | p. 133 |
4.7.2 RFS Representations of State and Detected Features | p. 135 |
4.7.3 Bayesian Formulation with a Finite Set Feature Map | p. 137 |
4.7.4 The Probability Hypothesis Density (PHD) Estimator | p. 138 |
4.7.5 The PHD Filter | p. 142 |
4.8 SLAM and FBRM Performance Metrics | p. 145 |
4.8.1 Vehicle State Estimate Evaluation | p. 145 |
4.8.2 Map Estimate Evaluation | p. 145 |
4.8.3 Evaluation of FBRM and SLAM with the Second Order Wasserstein Metric | p. 146 |
4.9 Summary | p. 148 |
4.10 Bibliographical Remarks | p. 149 |
4.10.1 Grid-Based Robotic Mapping (GBRM) | p. 149 |
4.10.2 Gaussian Approximations to Bayes Theorem | p. 150 |
4.10.3 Non-Parametric Approximations to Bayesian FB-SLAM | p. 152 |
4.10.4 Other Approximations to Bayesian FB-SLAM | p. 152 |
4.10.5 Feature Association and Management | p. 155 |
4.10.6 Random Finite Sets (RFSs) | p. 156 |
4.10.7 SLAM and FBRM Evaluation Metrics | p. 156 |
References | p. 157 |
Part II Radar Modeling and Scan Integration | p. 163 |
Chapter 5 Predicting and Simulating FMCW Radar Measurements | p. 165 |
5.1 Introduction | p. 165 |
5.2 FMCW Radar Detection in the Presence of Noise | p. 166 |
5.3 Noise Distributions During Target Absence and Presence | p. 168 |
5.3.1 Received Power Noise Estimation | p. 168 |
5.3.2 Range Noise Estimation | p. 169 |
5.4 Predicting Radar Measurements | p. 173 |
5.4.1 A-Scope Prediction Based on Expected Target RCS and Range | p. 173 |
5.4.2 A-Scope Prediction Based on Robot Motion | p. 174 |
5.5 Quantitative Comparison of Predicted and Actual Measurements | p. 176 |
5.6 A-scope Prediction Results | p. 177 |
5.6.1 Single Bearing A-Scope Prediction | p. 177 |
5.6.2 360° Scan Multiple A-Scope Prediction, Based on Robot Motion | p. 179 |
5.7 Summary | p. 188 |
5.8 Bibliographical Remarks | p. 192 |
References | p. 193 |
Chapter 6 Reducing Detection Errors and Noise with Multiple Radar Scans | p. 195 |
6.1 Introduction | p. 195 |
6.2 Radar Data in an Urban Environment | p. 196 |
6.2.1 Landmark Detection with Single Scan CA-CFAR | p. 198 |
6.3 Classical Scan Integration Methods | p. 198 |
6.3.1 Coherent and Noncoherent Integration | p. 198 |
6.3.2 Binary Integration Detection | p. 201 |
6.4 Integration Based on Target Presence Probability (TPP) Estimation | p. 204 |
6.5 False Alarm and Detection Probabilities for the TPP Estimator | p. 206 |
6.5.1 TPP Response to Noise: P TPP fa | p. 206 |
6.5.2 TPP Response to a Landmark and Noise: P TPP D | p. 209 |
6.5.3 Choice of ¿ p , T TTP (¿ p ,l) and l | p. 210 |
6.5.4 Numerical Method for Determining T TPP (¿ p , l ) and P TTP D | p. 211 |
6.6 A Comparison of Scan Integration Methods | p. 213 |
6.7 A Note on Multi-Path Reflections | p. 214 |
6.8 TPP Integration of Radar in an Urban Environment | p. 215 |
6.8.1 Qualitative Assessment of TPP Applied to A-Scope Information | p. 215 |
6.8.2 Quantitative Assessment of TPP Applied to Complete Scans | p. 215 |
6.8.3 A Qualitative Assessment of an Entire Parcking Lot Scene | p. 221 |
6.9 Recursive A-Scope Noise Reduction | p. 223 |
6.9.1 Single A-Scope Noise Subtraction | p. 225 |
6.9.2 Multiple A-Scope-Complete Scan Noise Subtraction | p. 227 |
6.10 Summary | p. 228 |
6.11 Bibliographical Remarks | p. 229 |
References | p. 230 |
Part III Robotic Mapping with Known Vehicle Location | p. 233 |
Chapter 7 Grid-Based Robotic Mapping with Detection Likelihood Filtering | p. 235 |
7.1 Introduction | p. 235 |
7.2 The Grid-Based Robotic Mapping (GBRM) Problem | p. 237 |
7.2.1 GBRM Based on Range Measurements | p. 239 |
7.2.2 GBRM with Detection Measurements | p. 241 |
7.2.3 Detection versus Range Measurement Models | p. 243 |
7.3 Mapping with Unknown Measurement Likelihoods | p. 245 |
7.3.1 Data Format | p. 245 |
7.3.2 GBRM Algorithm Overview | p. 246 |
7.3.3 Constant False Alarm Rate (CFAR) Detector | p. 247 |
7.3.4 Map Occupancy and Detection Likelihood Estimator | p. 247 |
7.3.5 Incorporation of the OS-CFAR Processor | p. 249 |
7.4 GBRM-ML Particle Filter Implementation | p. 250 |
7.5 Comparisons of Detection and Spatial-Based GBRM | p. 251 |
7.5.1 Dataset 1: Synthetic Data, Single Landmark | p. 251 |
7.5.2 Dataset 2: Real Experiments in the Parking Lot Environment | p. 252 |
7.5.3 Dataset 3: A Campus Environment | p. 259 |
7.6 Summary | p. 261 |
7.7 Bibliographical Remarks | p. 262 |
References | p. 263 |
Chapter 8 Feature-Based Robotic Mapping with Random Finite Sets | p. 267 |
8.1 Introduction | p. 267 |
8.2 The Probability Hypothesis Density (PHD)-FBRM Filter | p. 268 |
8.3 PHD-FBRM Filter Implementation | p. 269 |
8.3.1 The FBRM New Feature Proposal Strategy | p. 271 |
8.3.2 Gaussian Management and State Estimation | p. 272 |
8.3.3 GMM-PHD-FBRM Pseudo Code | p. 274 |
8.4 PHD-FBRM Computational Complexity | p. 275 |
8.5 Analysis of the PHD-FBRM Filter | p. 275 |
8.6 Summary | p. 279 |
8.7 Bibliographical Remarks | p. 280 |
References | p. 281 |
Part IV Simultaneous Localization and Mapping | p. 283 |
Chapter 9 Radar-Based SLAM with Random Finite Sets | p. 285 |
9.1 Introduction | p. 285 |
9.2 SLAM with the PHD Filter | p. 286 |
9.2.1 The Factorized RFS-SLAM Recursion | p. 286 |
9.2.2 PHD Mapping-Rao-Blackwellization | p. 287 |
9.2.3 PHD-SLAM | p. 288 |
9.3 Implementing the RB-PHD-SLAM Filter | p. 290 |
9.3.1 PHD Mapping-Implementation | p. 290 |
9.3.2 The Vehicle Trajectory-Implementation | p. 292 |
9.3.3 Estimating the Map | p. 292 |
9.3.4 GMM-PHD-SLAM Pseudo Code | p. 293 |
9.4 RB-PHD-SLAM Computational Complexity | p. 293 |
9.5 Radar-Based Comparisons of RFS and Vector-Based SLAM | p. 295 |
9.6 Summary | p. 299 |
9.7 Bibliographical Remarks | p. 299 |
References | p. 300 |
Chapter 10 X-Band Radar-Based SLAM in an Off-Shore Environment | p. 301 |
10.1 Introduction | p. 301 |
10.2 The ASC and the Coastal Environment | p. 303 |
10.3 Marine Radar Feature Extraction | p. 305 |
10.3.1 Adaptive Coastal Feature Detection-OS-CFAR | p. 306 |
10.3.2 Image-Based Smoothing-Gaussian Filtering | p. 308 |
10.3.3 Image-Based Thresholding | p. 311 |
10.3.4 Image-Based Clustering | p. 311 |
10.3.5 Feature Labeling | p. 313 |
10.4 The Marine Based SLAM Algorithms | p. 314 |
10.4.1 The ASC Process Model | p. 314 |
10.4.2 RFS SLAM with the PHD Filter | p. 314 |
10.4.3 NN-EKF-SLAM Implementation | p. 317 |
10.4.4 Multi-Hyphothesis (MH) FastSLAM Implementation | p. 317 |
10.5 Comparisons of SLAM Concepts at Sea | p. 318 |
10.5.1 SLAM Trial 1-Comparing PHD and NN-EKF-SLAM | p. 318 |
10.5.2 SLAM Trial 2-Comparing RB-PHD-SLAM and MH-FastSLAM | p. 322 |
10.6 Summary | p. 324 |
10.7 Bibliographical Remarks | p. 326 |
References | p. 326 |
Appendix A The Navtech FMCW MMW Radar Specifications | p. 329 |
Appendix B Derivation of g(Z k |Z k-1 ,X k ) for the RB-PHD-SLAM Filter | p. 331 |
B.1 The Empty Strategy | p. 331 |
B.2 The Single Feature Strategy | p. 332 |
Appendix c NN-EKF and FastSLAM Feature Management | p. 333 |
Index | p. 335 |