Sensor and Data Fusion for Intelligent Transportation Systems

Preface
List of Acronyms
1 Introduction
1.1 Applications to ITS
1.2 Data, Information, and Knowledge
1.3 Summary of Book Contents
References
2 Sensor and Data Fusion in Traffic Management
2.1 What is Meant by Sensor and Data Fusion?
2.2 Sensor and Data Fusion Benefits to Traffic Management
2.3 Data Sources for Traffic Management Applications
2.4 Sensor and Data Fusion Architectures
2.4.1 Architecture selection
2.4.2 Architecture classification
2.5 Detection, Classification, and Identification of a Vehicle
2.6 The JDL and DFIG Data Fusion Models
2.7 Level 1 Fusion: Detection, Classification, and Identification Algorithms
2.7.1 Physical models
2.7.2 Feature-based inference techniques
2.7.3 Cognitive-based models
2.8 Level 1 Fusion: State Estimation and Tracking Algorithms
2.8.1 Prediction gates, correlation metrics, and data association
2.8.2 Single- and two-level data and track association
2.8.3 Deterministic and probabilistic (all-neighbor) association
2.9 Data Fusion Algorithm Selection
2.10 Level 2 and Level 3 Fusion Processing
2.10.1 Level 2 processing
2.10.2 Level 3 processing
2.10.3 Situation awareness
2.10.4 Application to connected and self-driving vehicles
2.11 Level 4 Fusion Processing
2.12 Level 5 Fusion Processing
2.13 Applications of Sensor and Data Fusion to ITS
2.13.1 Advanced transportation management systems
2.13.2 Automatic incident detection
2.13.3 Network control
2.13.4 Advanced traveler information systems
2.13.5 Advanced driver assistance systems
2.13.6 Crash analysis and prevention
2.13.7 Traffic demand estimation
2.13.8 Traffic forecasting and traffic monitoring
2.13.9 Position and heading estimation
2.14 Summary
References
3 Bayesian Inference for Traffic Management
3.1 Bayesian Inference
3.2 Derivation of Bayes' Theorem
3.3 Likelihood Function and Prior Probability Models
3.4 Monty Hall Problem
3.4.1 Case-by-case analysis solution
3.4.2 Conditional probability solution
3.4.3 Bayesian inference solution
3.5 Application of Bayes' Theorem to Cancer Screening
3.6 Bayesian Inference in Support of Data Fusion
3.7 Bayesian Inference Applied to Vehicle Identification
3.8 Bayesian Inference Applied to Freeway Incident Detection Using Multiple Source Data
3.8.1 Problem development
3.8.2 Numerical example
3.9 Bayesian Inference Applied to Truck Classification
3.9.1 MCS architecture
3.9.2 MCS operation
3.9.3 Data collection and conclusions
3.10 Causal Bayesian Networks
3.10.1 Directed acyclic graphs
3.10.2 Application to maneuver-based trajectory prediction and criticality assessment
3.11 Summary
References
4 Dempster-Shafer Evidential Reasoning for Traffic Management
4.1 Overview of the Process
4.2 Implementation of the Method
4.3 Support, Plausibility, and Uncertainty Interval
4.4 Dempster's Rule for Combining Multiple-Sensor Data
4.5 Vehicle Detection Using Dempster-Shafer Evidential Reasoning
4.5.1 Dempster's rule applied to compatible data sets
4.5.2 Dempster's rule with null set elements
4.5.3 Dempster's rule with singleton propositions
4.6 Singleton Proposition Vehicle Detection Problem Solved with Bayesian Inference
4.7 Constructing Probability Mass Functions
4.7.1 Knowledge of sensor operation and object signature characteristics
4.7.2 Known probability distributions for the parameters of interest
4.7.3 Confusion matrix creation
4.7.4 Number and degree of matching of features to those of objects of interest
4.7.5 Exponential probability mass model
4.8 Decision Support System Application of Dempster-Shafer Reasoning
4.8.1 Field test description
4.8.2 Field test conclusions
4.9 Comparison with Bayesian Inference
4.10 Modifications to the Original Dempster-Shafer Method
4.11 Summary
References
5 Kalman Filtering for Traffic Management
5.1 Optimal Estimation
5.2 Kalman Filter Application to Object Tracking
5.3 State Transition Model
5.4 Measurement Model
5.4.1 Measurement error-covariance matrix for a three- and two-dimensional problem
5.4.2 Object in straight-line motion
5.5 The Discrete-Time Kalman Filter Algorithm
5.6 Relation of Measurement-to-Track Correlation Decision to the Kalman Gain
5.7 Initialization and Subsequent Recursive Operation of the Kalman Filter
5.8 ?-? Filter
5.8.1 Application and relation to Kalman gain
5.8.2 ?-? filter equations for state estimate prediction and correction
5.8.3 Noise reduction and transient response properties of the ?-? filter
5.8.4 Expressions for ? as a function of ?
5.9 Kalman Gain Control Methods
5.9.1 Preventing the gain from becoming too small
5.9.2 Preventing the gain from becoming too large
5.10 Noise Covariance Values and Filter Tuning
5.11 Process Noise Covariance Matrix Models
5.11.1 Constant velocity object process noise model
5.11.2 Constant acceleration object process noise model
5.12 Interacting Multiple Model for Vehicle Motion on a Roadway
5.12.1 Kinematic models
5.12.2 IMM implementation
5.12.3 Test results
5.13 Extended Kalman Filter
5.14 Summary
References
6 State of the Practice and Research Gaps
6.1 Data Fusion State of the Practice
6.2 Need for Continued Data Fusion Research
6.2.1 Reliability and quality of input data to the fusion system
6.2.2 Security of the data fusion system
6.2.3 Fusion of hard and soft data
6.2.4 Assessing the fusion system using measures of performance
6.2.5 Ground truth
6.2.6 Commercial database management system and operating system suitability
6.2.7 Design for worst-case data transmission and processing scenarios
6.2.8 Additional research needs
6.3 Prerequisite Information for Level 1 Object Assessment Algorithms
References
Appendix: The Fundamental Matrix of a Fixed Continuous-Time System
References
Index