Multimodal Perception and Secure State Estimation for Robotic Mobility Platforms

Xinghua Liu is a Professor with Xi'an University of Technology. His research interests are secure state estimation and control, cyber-physical systems, and artificial Intelligence.
 
Rui Jiang is a Staff Algorithm Engineer at the OmniVision Technologies Inc., and an Adjunct Lecturer with the National University of Singapore. His research interests are intelligent sensing, and perception for robotic systems.
 
Badong Chen is a Professor with Xi'an Jiaotong University. His research interests are signal processing, machine learning, artificial intelligence, neural engineering, and robotics.
 
Shuzhi Sam Ge is a Professor with the National University of Singapore and an honorary Director of Institute for Future, Qingdao University, China. His research interests are adaptive control, robotics, and artificial Intelligence.

Multimodal Perception and Secure State Estimation for Robotic Mobility Platforms
 
Enables readers to understand important new trends in multimodal perception for mobile robotics
 
This book provides a novel perspective on secure state estimation and multimodal perception for robotic mobility platforms such as autonomous vehicles. It thoroughly evaluates filter-based secure dynamic pose estimation approaches for autonomous vehicles over multiple attack signals and shows that they outperform conventional Kalman filtered results.
 
As a modern learning resource, it contains extensive simulative and experimental results that have been successfully implemented on various models and real platforms. To aid in reader comprehension, detailed and illustrative examples on algorithm implementation and performance evaluation are also presented. Written by four qualified authors in the field, sample topics covered in the book include:
* Secure state estimation that focuses on system robustness under cyber-attacks
* Multi-sensor fusion that helps improve system performance based on the complementary characteristics of different sensors
* A geometric pose estimation framework to incorporate measurements and constraints into a unified fusion scheme, which has been validated using public and self-collected data
* How to achieve real-time road-constrained and heading-assisted pose estimation
 
This book will appeal to graduate-level students and professionals in the fields of ground vehicle pose estimation and perception who are looking for modern and updated insight into key concepts related to the field of robotic mobility platforms.

About the Authors xii
 
Preface xiv
 
1 Introduction 1
 
1.1 Background and Motivation 1
 
1.2 Multimodal Pose Estimation for Vehicle Navigation 2
 
1.2.1 Multi-Senor Pose Estimation 2
 
1.2.2 Pose Estimation with Constraints 4
 
1.2.3 Research Focus in Multimodal Pose Estimation 5
 
1.3 Secure Estimation 7
 
1.3.1 Secure State Estimation under Cyber Attacks 7
 
1.3.2 Secure Pose Estimation for Autonomous Vehicles 8
 
1.4 Contributions and Organization 9
 
Part I Multimodal Perception in Vehicle Pose Estimation 13
 
2 Heading Reference-Assisted Pose Estimation 15
 
2.1 Preliminaries 16
 
2.1.1 Stereo Visual Odometry 16
 
2.1.2 Heading Reference Sensors 17
 
2.1.3 Graph Optimization on a Manifold 17
 
2.2 Abstraction Model of Measurement with a Heading Reference 19
 
2.2.1 Loosely Coupled Model 19
 
2.2.2 Tightly Coupled Model 20
 
2.2.3 Structure of the Abstraction Model 22
 
2.2.4 Vertex Removal in the Abstraction Model 22
 
2.3 Heading Reference-Assisted Pose Estimation (HRPE) 24
 
2.3.1 Initialization 24
 
2.3.2 Graph Optimization 24
 
2.3.3 Maintenance of the Dynamic Graph 26
 
2.4 Simulation Studies 26
 
2.4.1 Accuracy with Respect to Heading Measurement Error 28
 
2.4.2 Accuracy with Respect to Sliding Window Size 28
 
2.4.3 Time Consumption with Respect to Sliding Window Size 28
 
2.5 Experimental Results 31
 
2.5.1 Experimental Platform 31
 
2.5.2 Pose Estimation Performance 33
 
2.5.3 Real-Time Performance 34
 
2.6 Conclusion 36
 
3 Road-Constrained Localization Using Cloud Models 37
 
3.1 Preliminaries 38
 
3.1.1 Scaled Measurement Equations for Visual Odometry 38
 
3.1.2 Cloud Models 39
 
3.1.3 Uniform Gaussian Distribution (UGD) 39
 
3.1.4 Gaussian-Gaussian Distribution (GGD) 42
 
3.2 Map-Assisted Ground Vehicle Localization 43
 
3.2.1 Measurement Representation with UGD 44
 
3.2.2 Shape Matching Between Map and Particles 45
 
3.2.3 Particle Resampling and Parameter Estimation 46
 
3.2.4 Framework Extension to Other Cloud Models 47
 
3.3 Experimental Validation on UGD 47
 
3.3.1 Configurations 47
 
3.3.2 Localization with Stereo Visual Odometry 48
 
3.3.3 Localization with Monocular Visual Odometry 49
 
3.3.4 Scale Estimation Results 52
 
3.3.5 Weighting Function Balancing 52
 
3.4 Experimental Validation on GGD 54
 
3.4.1 Experiments on KITTI 55
 
3.4.2 Experiments on the Self-Collected Dataset 61
 
3.5 Conclusion 63
 
4 GPS/Odometry/Map Fusion for Vehicle Positioning Using Potential Functions 65
 
4.1 Potential Wells and Potential Trenches 66
 
4.1.1 Potential Function Creation 67
 
4.1.2 Minimum Searching 71
 
4.2 Potential-Function-Based Fusion for Vehicle Positioning 74
 
4.2.1 Information Sources and Sensors 74
 
4.2.2 Potential Representation 76
 
4.2.3 Road-Switching Strategy 76
 
4.3 Experimental Results 78
 
4.3.1 Quantitative Results 78
 
4.3.2 Qualitative Evaluation 80
 
4.4 Conclusion 84
 
5 Multi-Sensor Geometric Pose Estimation 85
 
5.1 Preliminaries 86
 
5.1.1 Distance on Riemannian Manifolds 86
 
5.1.2 Probabilistic Distribution on Riemannian Manifolds 87
 
5.2 Geometric Pose Estimation Using Dynamic Potential Fields 88
 
5.2.1 State Space and Measurement Space 88
 
5.2.2 Dynamic Potential