Cover image for Simultaneous localization and mapping : exactly sparse information filters
Title:
Simultaneous localization and mapping : exactly sparse information filters
Personal Author:
Series:
New frontiers in robotics ; 3
Physical Description:
xii, 194 pages : illustrations ; 23 cm.
ISBN:
9789814350310

Available:*

Library
Item Barcode
Call Number
Material Type
Item Category 1
Status
Searching...
30000010280455 TJ211.415 W365 2011 Open Access Book Book
Searching...

On Order

Summary

Summary

Simultaneous localization and mapping (SLAM) is a process where an autonomous vehicle builds a map of an unknown environment while concurrently generating an estimate for its location. This book is concerned with computationally efficient solutions to the large scale SLAM problems using exactly sparse Extended Information Filters (EIF).
The invaluable book also provides a comprehensive theoretical analysis of the properties of the information matrix in EIF-based algorithms for SLAM. Three exactly sparse information filters for SLAM are described in detail, together with two efficient and exact methods for recovering the state vector and the covariance matrix. Proposed algorithms are extensively evaluated both in simulation and through experiments.


Table of Contents

Prefacep. v
Acknowledgmentsp. vii
1 Introductionp. 1
1.1 The SLAM Problem and Its Applicationsp. 2
1.1.1 Description of the SLAM Problemp. 2
1.1.2 Applications of SLAMp. 3
1.2 Summary of SLAM Approachesp. 4
1.2.1 EKF/EIF based SLAM Approachesp. 5
1.2.2 Other SLAM Approachesp. 8
1.3 Key Properties of SLAMp. 12
1.3.1 Observabilityp. 12
1.3.2 EKF SLAM Convergencep. 19
1.3.3 EKF SLAM Consistencyp. 34
1.4 Motivationp. 41
1.5 Book Overviewp. 43
2 Sparse Information Filters in SLAMp. 47
2.1 Information Matrix in the Full SLAM Formulationp. 47
2.2 Information Matrix in the Conventional EIF SLAM Formulationp. 52
2.3 Meaning of Zero Off-diagonal Elements in Information Matrixp. 54
2.4 Conditions for Achieving Exact Sparsenessp. 57
2.5 Strategies for Achieving Exact Sparsenessp. 59
2.5.1 Decoupling Localization and Mappingp. 59
2.5.2 Using Local Submapsp. 60
2.5.3 Combining Decoupling and Submapsp. 60
2.6 Important Practical Issues in EIF SLAMp. 61
2.7 Summaryp. 62
3 Decoupling Localization and Mappingp. 63
3.1 The D-SLAM Algorithmp. 64
3.1.1 Extracting Map Information from Observationsp. 64
3.1.2 Key Idea of D-SLAMp. 69
3.1.3 Mappingp. 69
3.1.4 Localizationp. 71
3.2 Structure of the Information Matrix in D-SLAMp. 77
3.3 Efficient State and Covariance Recoverp. 78
3.3.1 Recovery Using the Preconditioned Conjugated Gradient (PCG) Methodp. 80
3.3.2 Recovery Using Complete Cholesky Factorizationp. 82
3.4 Implementation Issuesp. 84
3.4.1 Admissible Measurementsp. 84
3.4.2 Data Associationp. 86
3.5 Computer Simulationsp. 88
3.6 Experimental Evaluationp. 95
3.6.1 Experiment in a Small Environmentp. 95
3.6.2 Experiment Using the Victoria Park Datasetp. 95
3.7 Computational Complexityp. 99
3.7.1 Storagep. 102
3.7.2 Localizationp. 102
3.7.3 Mapping
3.7.4 State and Covariance Recoveryp. 103
3.8 Consistency of D-SLAMp. 107
3.9 Bibliographical Remarksp. 108
3.10 Summaryp. 111
4 D-SLAM Local Map Joining Filterp. 113
4.1 Structure of D-SLAM Local Map Joining Filterp. 114
4.1.1 State Vectorsp. 115
4.1.2 Relative Information Relating Feature Locationsp. 116
4.1.3 Combining Local Maps Using Relative Informationp. 116
4.2 Obtaining Relative Location Information in Local Maps
4.2.1 Generating a Local Mapp. 117
4.2.2 Obtaining Relative Location Information in the Local Mapp. 118
4.3 Global Map Updatep. 122
4.3.1 Measurement Modelp. 122
4.3.2 Updating the Global Mapp. 122
4.3.3 Sparse Information Matrixp. 124
4.4 Implementation Issuesp. 125
4.4.1 Robot Localizationp. 125
4.4.2 Data Associationp. 126
4.4.3 State and Covariance Recoveryp. 127
4.4.4 When to Start a New Local Mapp. 128
4.5 Computational Complexityp. 128
4.5.1 Storagep. 128
4.5.2 Local Map Constructionp. 129
4.5.3 Global Map Updatep. 129
4.5.4 Rescheduling the Computational Effortp. 130
4.6 Computer Simulationsp. 130
4.6.1 Simulation in a Small Areap. 130
4.6.2 Simulation in a Large Areap. 134
4.7 Experimental Evaluationp. 140
4.8 Bibliographical Remarksp. 147
4.9 Summaryp. 149
5 Sparse Local Submap Joining Filterp. 151
5.1 Structure of Sparse Local Submap Joining Filterp. 152
5.1.1 Input to SLSJF - Local Mapsp. 153
5.1.2 Output of SLSJF - One Global Mapp. 154
5.2 Fusing Local Maps into the Global Mapp. 155
5.2.1 Adding X (K+1)s G into the Global Mapp. 155
5.2.2 Initializing the Values of New Features and (K+1)e G in the Global Mapp. 156
5.2.3 Updating the Global Mapp. 157
5.3 Sparse Information Matrixp. 158
5.4 Implementation Issuesp. 159
5.4.1 Data Associationp. 160
5.4.2 State and Covariance Recoveryp. 162
5.5 Computer Simulationsp. 162
5.6 Experimental Evaluationp. 169
5.7 Discussionp. l69
5.7.1 Computational Complexityp. 169
5.7.2 Zero Information Lossp. 173
5 7.3 Tradeoffs in Achieving Exactly Sparse Representationp. 174
5.8 Summaryp. 175
Appendix A Proofs of EKF SLAM Convergence and Consistencyp. 177
A.l Matrix Inversion Lemmap. 177
A.2 Proofs of EKF SLAM Convergencep. 178
A.3 Proofs of EKF SLAM Consistencyp. 181
Appendix B Incremental Method for Cholesky Factorization of SLAM Information Matrixp. 185
B.l Cholesky Factorizationp. 185
B.2 Approximate Cholesky Factorizationp. 186
Bibliographyp. 189