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.

• About the Authors xiiPreface xiv1 Introduction 11.1 Background and Motivation 11.2 Multimodal Pose Estimation for Vehicle Navigation 21.2.1 Multi-Senor Pose Estimation 21.2.2 Pose Estimation with Constraints 41.2.3 Research Focus in Multimodal Pose Estimation 51.3 Secure Estimation 71.3.1 Secure State Estimation under Cyber Attacks 71.3.2 Secure Pose Estimation for Autonomous Vehicles 81.4 Contributions and Organization 9Part I Multimodal Perception in Vehicle Pose Estimation 132 Heading Reference-Assisted Pose Estimation 152.1 Preliminaries 162.1.1 Stereo Visual Odometry 162.1.2 Heading Reference Sensors 172.1.3 Graph Optimization on a Manifold 172.2 Abstraction Model of Measurement with a Heading Reference 192.2.1 Loosely Coupled Model 192.2.2 Tightly Coupled Model 202.2.3 Structure of the Abstraction Model 222.2.4 Vertex Removal in the Abstraction Model 222.3 Heading Reference-Assisted Pose Estimation (HRPE) 242.3.1 Initialization 242.3.2 Graph Optimization 242.3.3 Maintenance of the Dynamic Graph 262.4 Simulation Studies 262.4.1 Accuracy with Respect to Heading Measurement Error 282.4.2 Accuracy with Respect to Sliding Window Size 282.4.3 Time Consumption with Respect to Sliding Window Size 282.5 Experimental Results 312.5.1 Experimental Platform 312.5.2 Pose Estimation Performance 332.5.3 Real-Time Performance 342.6 Conclusion 363 Road-Constrained Localization Using Cloud Models 373.1 Preliminaries 383.1.1 Scaled Measurement Equations for Visual Odometry 383.1.2 Cloud Models 393.1.3 Uniform Gaussian Distribution (UGD) 393.1.4 Gaussian-Gaussian Distribution (GGD) 423.2 Map-Assisted Ground Vehicle Localization 433.2.1 Measurement Representation with UGD 443.2.2 Shape Matching Between Map and Particles 453.2.3 Particle Resampling and Parameter Estimation 463.2.4 Framework Extension to Other Cloud Models 473.3 Experimental Validation on UGD 473.3.1 Configurations 473.3.2 Localization with Stereo Visual Odometry 483.3.3 Localization with Monocular Visual Odometry 493.3.4 Scale Estimation Results 523.3.5 Weighting Function Balancing 523.4 Experimental Validation on GGD 543.4.1 Experiments on KITTI 553.4.2 Experiments on the Self-Collected Dataset 613.5 Conclusion 634 GPS/Odometry/Map Fusion for Vehicle Positioning Using Potential Functions 654.1 Potential Wells and Potential Trenches 664.1.1 Potential Function Creation 674.1.2 Minimum Searching 714.2 Potential-Function-Based Fusion for Vehicle Positioning 744.2.1 Information Sources and Sensors 744.2.2 Potential Representation 764.2.3 Road-Switching Strategy 764.3 Experimental Results 784.3.1 Quantitative Results 784.3.2 Qualitative Evaluation 804.4 Conclusion 845 Multi-Sensor Geometric Pose Estimation 855.1 Preliminaries 865.1.1 Distance on Riemannian Manifolds 865.1.2 Probabilistic Distribution on Riemannian Manifolds 875.2 Geometric Pose Estimation Using Dynamic Potential Fields 885.2.1 State Space and Measurement Space 885.2.2 Dynamic Potential Fields on Manifolds 905.2.3 DPF-Based Information Fusion 915.2.4 Approximation of Geometric Pose Estimation 955.3 VO-Heading-Map Pose Estimation for Ground Vehicles 975.3.1 System Modeling 975.3.2 Road Constraints 985.3.3 Parameter Estimation on SE(3) 995.4 Experiments on KITTI Sequences 995.4.1 Overall Performance 995.4.2 Influence of Heading Error 1025.4.3 Influence of Road Map Resolution 1025.4.4 Influences of Parameters 1045.5 Experiments on the NTU Dataset 1055.5.1 Overall Performance 1055.5.2 Phenomena Observed During Experiments 1055.6 Conclusion 107Part II Secure State Estimation for Mobile Robots 1096 Filter-Based Secure Dynamic Pose Estimation 1116.1 Introduction 1116.2 RelatedWork 1136.3 Problem Formulation 1146.3.1 System Model 1146.3.2 Measurement Model 1166.3.3 Attack Model 1166.4 Estimator Design 1176.5 Discussion of Parameter Selection 1226.5.1 The Probability Subject to Deception Attacks 1226.5.2 The Bound of Signal ;;k 1236.6 Experimental Validation 1236.6.1 Pose Estimation under Attack on a Single State 1256.6.2 Pose Estimation under Attacks on Multiple States 1276.7 Conclusion 1307 UKF-Based Vehicle Pose Estimation under Randomly Occurring Deception Attacks 1317.1 Introduction 1317.2 Related Work 1337.3 Pose Estimation Problem for Ground Vehicles under Attack 1347.3.1 System Model 1347.3.2 Attack Model 1367.4 Design of the Unscented Kalman Filter 1377.5 Numeric Simulation 1417.6 Experiments 1447.6.1 General Performance 1457.6.2 Influence of Parameters 1457.7 Conclusion 1478 Secure Dynamic State Estimation with a Decomposing Kalman Filter 1498.1 Introduction 1498.2 Problem Formulation 1518.3 Decomposition of the Kalman Filter By Using a Local Estimate 1538.4 A Secure Information Fusion Scheme 1588.5 Numerical Example 1618.6 Conclusion 1628.7 Appendix: Proof of Theorem 8.2 1628.8 Proof of Theorem 8.4 1659 Secure Dynamic State Estimation for AHRS 1699.1 Introduction 1699.2 Related Work 1709.2.1 Attitude Estimation 1709.2.2 Secure State Estimation 1719.2.3 Secure Attitude Estimation 1719.3 Attitude Estimation Using Heading References 1729.3.1 Attitude Estimation from Vector Observations 1729.3.2 Secure Attitude Estimation Framework and Modeling 1739.4 Secure Estimator Design with a Decomposing Kalman Filter 1749.4.1 Decomposition of the Kalman Filter Using a Local Estimate 1769.4.2 A Least-Square Interpretation for the Decomposition 1779.4.3 Secure State Estimate 1789.5 Simulation Validation 1819.5.1 Simulating Measurements with Attacks 1829.5.2 Filter Performance 1829.5.3 Influence of Parameter ;; 1829.6 Conclusion 18410 Conclusions 185References 189Index 207

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.

• About the Authors xiiPreface xiv1 Introduction 11.1 Background and Motivation 11.2 Multimodal Pose Estimation for Vehicle Navigation 21.2.1 Multi-Senor Pose Estimation 21.2.2 Pose Estimation with Constraints 41.2.3 Research Focus in Multimodal Pose Estimation 51.3 Secure Estimation 71.3.1 Secure State Estimation under Cyber Attacks 71.3.2 Secure Pose Estimation for Autonomous Vehicles 81.4 Contributions and Organization 9Part I Multimodal Perception in Vehicle Pose Estimation 132 Heading Reference-Assisted Pose Estimation 152.1 Preliminaries 162.1.1 Stereo Visual Odometry 162.1.2 Heading Reference Sensors 172.1.3 Graph Optimization on a Manifold 172.2 Abstraction Model of Measurement with a Heading Reference 192.2.1 Loosely Coupled Model 192.2.2 Tightly Coupled Model 202.2.3 Structure of the Abstraction Model 222.2.4 Vertex Removal in the Abstraction Model 222.3 Heading Reference-Assisted Pose Estimation (HRPE) 242.3.1 Initialization 242.3.2 Graph Optimization 242.3.3 Maintenance of the Dynamic Graph 262.4 Simulation Studies 262.4.1 Accuracy with Respect to Heading Measurement Error 282.4.2 Accuracy with Respect to Sliding Window Size 282.4.3 Time Consumption with Respect to Sliding Window Size 282.5 Experimental Results 312.5.1 Experimental Platform 312.5.2 Pose Estimation Performance 332.5.3 Real-Time Performance 342.6 Conclusion 363 Road-Constrained Localization Using Cloud Models 373.1 Preliminaries 383.1.1 Scaled Measurement Equations for Visual Odometry 383.1.2 Cloud Models 393.1.3 Uniform Gaussian Distribution (UGD) 393.1.4 Gaussian-Gaussian Distribution (GGD) 423.2 Map-Assisted Ground Vehicle Localization 433.2.1 Measurement Representation with UGD 443.2.2 Shape Matching Between Map and Particles 453.2.3 Particle Resampling and Parameter Estimation 463.2.4 Framework Extension to Other Cloud Models 473.3 Experimental Validation on UGD 473.3.1 Configurations 473.3.2 Localization with Stereo Visual Odometry 483.3.3 Localization with Monocular Visual Odometry 493.3.4 Scale Estimation Results 523.3.5 Weighting Function Balancing 523.4 Experimental Validation on GGD 543.4.1 Experiments on KITTI 553.4.2 Experiments on the Self-Collected Dataset 613.5 Conclusion 634 GPS/Odometry/Map Fusion for Vehicle Positioning Using Potential Functions 654.1 Potential Wells and Potential Trenches 664.1.1 Potential Function Creation 674.1.2 Minimum Searching 714.2 Potential-Function-Based Fusion for Vehicle Positioning 744.2.1 Information Sources and Sensors 744.2.2 Potential Representation 764.2.3 Road-Switching Strategy 764.3 Experimental Results 784.3.1 Quantitative Results 784.3.2 Qualitative Evaluation 804.4 Conclusion 845 Multi-Sensor Geometric Pose Estimation 855.1 Preliminaries 865.1.1 Distance on Riemannian Manifolds 865.1.2 Probabilistic Distribution on Riemannian Manifolds 875.2 Geometric Pose Estimation Using Dynamic Potential Fields 885.2.1 State Space and Measurement Space 885.2.2 Dynamic Potential Fields on Manifolds 905.2.3 DPF-Based Information Fusion 915.2.4 Approximation of Geometric Pose Estimation 955.3 VO-Heading-Map Pose Estimation for Ground Vehicles 975.3.1 System Modeling 975.3.2 Road Constraints 985.3.3 Parameter Estimation on SE(3) 995.4 Experiments on KITTI Sequences 995.4.1 Overall Performance 995.4.2 Influence of Heading Error 1025.4.3 Influence of Road Map Resolution 1025.4.4 Influences of Parameters 1045.5 Experiments on the NTU Dataset 1055.5.1 Overall Performance 1055.5.2 Phenomena Observed During Experiments 1055.6 Conclusion 107Part II Secure State Estimation for Mobile Robots 1096 Filter-Based Secure Dynamic Pose Estimation 1116.1 Introduction 1116.2 RelatedWork 1136.3 Problem Formulation 1146.3.1 System Model 1146.3.2 Measurement Model 1166.3.3 Attack Model 1166.4 Estimator Design 1176.5 Discussion of Parameter Selection 1226.5.1 The Probability Subject to Deception Attacks 1226.5.2 The Bound of Signal ;;k 1236.6 Experimental Validation 1236.6.1 Pose Estimation under Attack on a Single State 1256.6.2 Pose Estimation under Attacks on Multiple States 1276.7 Conclusion 1307 UKF-Based Vehicle Pose Estimation under Randomly Occurring Deception Attacks 1317.1 Introduction 1317.2 Related Work 1337.3 Pose Estimation Problem for Ground Vehicles under Attack 1347.3.1 System Model 1347.3.2 Attack Model 1367.4 Design of the Unscented Kalman Filter 1377.5 Numeric Simulation 1417.6 Experiments 1447.6.1 General Performance 1457.6.2 Influence of Parameters 1457.7 Conclusion 1478 Secure Dynamic State Estimation with a Decomposing Kalman Filter 1498.1 Introduction 1498.2 Problem Formulation 1518.3 Decomposition of the Kalman Filter By Using a Local Estimate 1538.4 A Secure Information Fusion Scheme 1588.5 Numerical Example 1618.6 Conclusion 1628.7 Appendix: Proof of Theorem 8.2 1628.8 Proof of Theorem 8.4 1659 Secure Dynamic State Estimation for AHRS 1699.1 Introduction 1699.2 Related Work 1709.2.1 Attitude Estimation 1709.2.2 Secure State Estimation 1719.2.3 Secure Attitude Estimation 1719.3 Attitude Estimation Using Heading References 1729.3.1 Attitude Estimation from Vector Observations 1729.3.2 Secure Attitude Estimation Framework and Modeling 1739.4 Secure Estimator Design with a Decomposing Kalman Filter 1749.4.1 Decomposition of the Kalman Filter Using a Local Estimate 1769.4.2 A Least-Square Interpretation for the Decomposition 1779.4.3 Secure State Estimate 1789.5 Simulation Validation 1819.5.1 Simulating Measurements with Attacks 1829.5.2 Filter Performance 1829.5.3 Influence of Parameter ;; 1829.6 Conclusion 18410 Conclusions 185References 189Index 207

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.

• About the Authors xiiPreface xiv1 Introduction 11.1 Background and Motivation 11.2 Multimodal Pose Estimation for Vehicle Navigation 21.2.1 Multi-Senor Pose Estimation 21.2.2 Pose Estimation with Constraints 41.2.3 Research Focus in Multimodal Pose Estimation 51.3 Secure Estimation 71.3.1 Secure State Estimation under Cyber Attacks 71.3.2 Secure Pose Estimation for Autonomous Vehicles 81.4 Contributions and Organization 9Part I Multimodal Perception in Vehicle Pose Estimation 132 Heading Reference-Assisted Pose Estimation 152.1 Preliminaries 162.1.1 Stereo Visual Odometry 162.1.2 Heading Reference Sensors 172.1.3 Graph Optimization on a Manifold 172.2 Abstraction Model of Measurement with a Heading Reference 192.2.1 Loosely Coupled Model 192.2.2 Tightly Coupled Model 202.2.3 Structure of the Abstraction Model 222.2.4 Vertex Removal in the Abstraction Model 222.3 Heading Reference-Assisted Pose Estimation (HRPE) 242.3.1 Initialization 242.3.2 Graph Optimization 242.3.3 Maintenance of the Dynamic Graph 262.4 Simulation Studies 262.4.1 Accuracy with Respect to Heading Measurement Error 282.4.2 Accuracy with Respect to Sliding Window Size 282.4.3 Time Consumption with Respect to Sliding Window Size 282.5 Experimental Results 312.5.1 Experimental Platform 312.5.2 Pose Estimation Performance 332.5.3 Real-Time Performance 342.6 Conclusion 363 Road-Constrained Localization Using Cloud Models 373.1 Preliminaries 383.1.1 Scaled Measurement Equations for Visual Odometry 383.1.2 Cloud Models 393.1.3 Uniform Gaussian Distribution (UGD) 393.1.4 Gaussian-Gaussian Distribution (GGD) 423.2 Map-Assisted Ground Vehicle Localization 433.2.1 Measurement Representation with UGD 443.2.2 Shape Matching Between Map and Particles 453.2.3 Particle Resampling and Parameter Estimation 463.2.4 Framework Extension to Other Cloud Models 473.3 Experimental Validation on UGD 473.3.1 Configurations 473.3.2 Localization with Stereo Visual Odometry 483.3.3 Localization with Monocular Visual Odometry 493.3.4 Scale Estimation Results 523.3.5 Weighting Function Balancing 523.4 Experimental Validation on GGD 543.4.1 Experiments on KITTI 553.4.2 Experiments on the Self-Collected Dataset 613.5 Conclusion 634 GPS/Odometry/Map Fusion for Vehicle Positioning Using Potential Functions 654.1 Potential Wells and Potential Trenches 664.1.1 Potential Function Creation 674.1.2 Minimum Searching 714.2 Potential-Function-Based Fusion for Vehicle Positioning 744.2.1 Information Sources and Sensors 744.2.2 Potential Representation 764.2.3 Road-Switching Strategy 764.3 Experimental Results 784.3.1 Quantitative Results 784.3.2 Qualitative Evaluation 804.4 Conclusion 845 Multi-Sensor Geometric Pose Estimation 855.1 Preliminaries 865.1.1 Distance on Riemannian Manifolds 865.1.2 Probabilistic Distribution on Riemannian Manifolds 875.2 Geometric Pose Estimation Using Dynamic Potential Fields 885.2.1 State Space and Measurement Space 885.2.2 Dynamic Potential Fields on Manifolds 905.2.3 DPF-Based Information Fusion 915.2.4 Approximation of Geometric Pose Estimation 955.3 VO-Heading-Map Pose Estimation for Ground Vehicles 975.3.1 System Modeling 975.3.2 Road Constraints 985.3.3 Parameter Estimation on SE(3) 995.4 Experiments on KITTI Sequences 995.4.1 Overall Performance 995.4.2 Influence of Heading Error 1025.4.3 Influence of Road Map Resolution 1025.4.4 Influences of Parameters 1045.5 Experiments on the NTU Dataset 1055.5.1 Overall Performance 1055.5.2 Phenomena Observed During Experiments 1055.6 Conclusion 107Part II Secure State Estimation for Mobile Robots 1096 Filter-Based Secure Dynamic Pose Estimation 1116.1 Introduction 1116.2 RelatedWork 1136.3 Problem Formulation 1146.3.1 System Model 1146.3.2 Measurement Model 1166.3.3 Attack Model 1166.4 Estimator Design 1176.5 Discussion of Parameter Selection 1226.5.1 The Probability Subject to Deception Attacks 1226.5.2 The Bound of Signal ;;k 1236.6 Experimental Validation 1236.6.1 Pose Estimation under Attack on a Single State 1256.6.2 Pose Estimation under Attacks on Multiple States 1276.7 Conclusion 1307 UKF-Based Vehicle Pose Estimation under Randomly Occurring Deception Attacks 1317.1 Introduction 1317.2 Related Work 1337.3 Pose Estimation Problem for Ground Vehicles under Attack 1347.3.1 System Model 1347.3.2 Attack Model 1367.4 Design of the Unscented Kalman Filter 1377.5 Numeric Simulation 1417.6 Experiments 1447.6.1 General Performance 1457.6.2 Influence of Parameters 1457.7 Conclusion 1478 Secure Dynamic State Estimation with a Decomposing Kalman Filter 1498.1 Introduction 1498.2 Problem Formulation 1518.3 Decomposition of the Kalman Filter By Using a Local Estimate 1538.4 A Secure Information Fusion Scheme 1588.5 Numerical Example 1618.6 Conclusion 1628.7 Appendix: Proof of Theorem 8.2 1628.8 Proof of Theorem 8.4 1659 Secure Dynamic State Estimation for AHRS 1699.1 Introduction 1699.2 Related Work 1709.2.1 Attitude Estimation 1709.2.2 Secure State Estimation 1719.2.3 Secure Attitude Estimation 1719.3 Attitude Estimation Using Heading References 1729.3.1 Attitude Estimation from Vector Observations 1729.3.2 Secure Attitude Estimation Framework and Modeling 1739.4 Secure Estimator Design with a Decomposing Kalman Filter 1749.4.1 Decomposition of the Kalman Filter Using a Local Estimate 1769.4.2 A Least-Square Interpretation for the Decomposition 1779.4.3 Secure State Estimate 1789.5 Simulation Validation 1819.5.1 Simulating Measurements with Attacks 1829.5.2 Filter Performance 1829.5.3 Influence of Parameter ;; 1829.6 Conclusion 18410 Conclusions 185References 189Index 207