Machine Learning and Optimization Techniques for Automotive Cyber-Physical Systems

✍ Scribed by Vipin Kumar Kukkala; Sudeep Pasricha

• Publisher: Springer Nature
• Year: 2023
• Tongue: English
• Leaves: 782
• Category: Library

✦ Synopsis

This book provides comprehensive coverage of various solutions that address issues related to real-time performance, security, and robustness in emerging automotive platforms. The authors discuss recent advances towards the goal of enabling reliable, secure, and robust, time-critical automotive cyber-physical systems, using advanced optimization and machine learning techniques. The focus is on presenting state-of-the-art solutions to various challenges including real-time data scheduling, secure communication within and outside the vehicle, tolerance to faults, optimizing the use of resource-constrained automotive ECUs, intrusion detection, and developing robust perception and control techniques for increasingly autonomous vehicles.

✦ Table of Contents

Preface
Acknowledgments
Contents
Part I Real-Time Scheduling
Reliable Real-Time Message Scheduling in Automotive Cyber-Physical Systems
1 Introduction
2 FlexRay Overview
3 Related Work
4 Problem Definition
4.1 System Model
4.2 Jitter Model
4.3 Hybrid SA+GRASP Heuristic
4.3.1 Simulated Annealing
4.3.2 Greedy Randomized Adaptive Search Procedure
4.3.3 Hybrid Heuristic Formulation
4.4 Inputs and Definitions
5 JAMS-SG Framework Overview
5.1 Jitter-Aware Design Time Frame Packing
5.1.1 Greedy Randomized Construction
5.1.2 Local Search
5.1.3 Design Time Scheduling
5.1.4 Acceptance and Cooling Functions
5.2 Runtime Multi-level Feedback Queue
5.3 Runtime Scheduler
6 Experiments
6.1 Experimental Setup
6.2 Comparison of JAMS-SG Variants
6.3 Response Time Analysis
6.4 Sensitivity Analysis
6.5 Scalability Analysis
7 Conclusion
References
Evolvement of Scheduling Theories for Autonomous Vehicles
1 Introduction
1.1 Organization
2 Background
2.1 Scheduling and Analyzing DAG Tasks in Autonomous Vehicles
2.1.1 The State-of-the-Art in DAG Scheduling and Analysis
2.2 Real-Time Scheduling for Reliable Autonomous Driving
2.2.1 Fault Tolerance
2.2.2 Resource Sharing
2.2.3 Mixed Criticality System
2.3 Real-Time TSN Scheduling for Automotive CPS
3 Scheduling of DAGs on Multiprocessor Architectures
3.1 Task Model and Scheduling Preliminaries
3.2 Concurrent Provider and Consumer Model
3.3 DAG Scheduling: A Parallelism and Dependency Exploited Method
3.3.1 The ``Critical Path First'' Execution (CPFE)
3.3.2 Exploiting Parallelism and Node Dependency
4 Reliable Resource Sharing in Reliable Autonomous Driving
4.1 System and Task Model
4.2 A Fault-Tolerant Solution for MCS with Shared Resources
4.2.1 The Fault-Tolerance System Model
4.2.2 Fault-Tolerance of Normal Sections
4.2.3 Fault-Tolerance of Critical Sections by MSRP-FT
5 Real-Time TSN Scheduling for Automotive CPS
5.1 Overview of Traffic Scheduling of TSN
5.2 Scheduling Network Packets in TSN
5.3 Deferred Queue
5.4 Worst-Case Response Time Analysis
5.5 Controller Synthesis and Period Allocation
5.5.1 Control Model
5.5.2 Problem Definition
5.5.3 Solving the Network and Control Co-Design Problem
6 Conclusion
References
Distributed Coordination and Centralized Scheduling for Automobiles at Intersections
1 Introduction
2 Distributed Coordination
2.1 Problem Formulation
2.1.1 Assumption on Fixed Paths
2.1.2 Notations of Discrete States
2.2 Distributed Coordination Approach
2.2.1 Decision Making
2.2.2 Motion Planning under Temporal Constraints
2.2.3 Theoretical Guarantees
2.3 Simulation Results
2.3.1 Microscopic Case Study
2.3.2 Macroscopic Traffic Simulation
2.4 Conclusion
3 Centralized Scheduling
3.1 Problem Formulation
3.2 Deadlock-Freeness Verification
3.3 Centralized Scheduling Approach
3.3.1 Definitions
3.3.2 Cycle-Removal-Based Scheduling
3.4 A Special Case: Lane Merging
3.5 Experimental Results
3.5.1 Scheduling Effectiveness and Efficiency
3.5.2 Modeling Expressiveness
3.6 Conclusion
4 Summary
References
Part II Security-Aware Design
Security-Aware Design of Time-Critical Automotive Cyber-Physical Systems
1 Introduction
2 Related Work
3 Problem Definition
3.1 System and Application Model
3.2 FlexRay Communication Protocol
3.3 Attack Models
3.4 Security Model
3.5 Definitions
4 SEDAN Framework: Overview
4.1 Task Allocation
4.2 Frame Packing
4.3 Deriving Security Requirements
4.4 Optimizing Message Key Sizes Using GRASP
4.4.1 Greedy Randomized Construction Phase
4.4.2 Local Search Phase
4.5 Setting Up Session Key
4.6 Authenticated Encryption/Decryption
4.7 Runtime Message Scheduler
5 Experiments
5.1 Experimental Setup
5.2 Benchmarking Encryption Algorithms
5.3 GRASP Parameter Selection
5.4 Response Time Analysis
5.5 Security Analysis
6 Conclusions
References
Secure by Design Autonomous Emergency Braking Systems in Accordance with ISO 21434
1 Introduction
2 Background
2.1 The AEB System in a Nutshell
2.2 Overview of ISO 21434 Activities
3 In-Depth Analysis of Adversarial Actions on AEB Control Systems
3.1 Detailed AEB System Model
3.2 Adversary Model and Attack Strategies
3.3 Attack Evaluation on the AEB Model
3.4 Impact of Stealthy Attacks
4 Secure-by-Design AEB in Accordance to ISO 21434
4.1 Overview of the ISO 21434 Cybersecurity Design Flow
4.2 From Item Definition to Risk Determination
4.3 From Determined Risks to Cybersecurity Goals and Concept
5 Conclusion
References
Resource Aware Synthesis of Automotive Security Primitives
1 Introduction
2 Background and Related Work
2.1 System Model for Secure CPS
2.2 Automotive Software Tools and Standards
2.3 Related Studies
3 Lightweight Attack Detection
3.1 Optimal Static Threshold-Based Detector
3.2 Variable Threshold-Based Attack Detector
3.2.1 Pivot-Based Threshold Synthesis Method
3.2.2 Step-Wise Threshold Synthesis Method
4 AI-Based Adaptive Attack Detection
4.1 Optimal Attack Policy Design
4.2 The MARL Based Framework
5 Attack Mitigation
6 A Platform Level Example
7 Conclusion
References
Gradient-Free Adversarial Attacks on 3D Point Clouds from LiDAR Sensors
1 Introduction
1.1 Contributions
1.2 Outline
2 Background
2.1 Semantic Segmentation of LiDAR Data
2.2 Existing Attacks on LiDAR Sensors
2.3 Adversarial Attacks
2.3.1 Attack Goal
2.3.2 Known Information of the Attacker
2.3.3 Adversarial Training
2.4 Adversarial Attacks on LiDAR Point Clouds
3 Threat Model
4 Gradient-Free Adversarial Attacks on Semantic Segmentation of LiDAR Data
4.1 Adversarial Attack as Multi-Objective Optimization Problem
4.2 Evolutionary Algorithms for Multi-Objective Optimization
4.3 Gradient-Free Adversarial Attack Methodology
4.4 Encoding
4.4.1 Manipulation of Arbitrary 3D Points
4.4.2 Manipulation of 3D Points within Cuboid
4.5 Evaluation
4.6 Implementation Details
5 Experiment Results
5.1 Hypotheses and Focus of Experiments
5.2 Run Time and Effectiveness
5.3 Manipulation of Arbitrary 3D Points
5.4 Manipulation of 3D Points Within Cuboid
5.5 Attacking Class Labels of 3D Points Within an RoI
5.6 Interpretation of Results
6 Conclusion and Future Work
6.1 Conclusion
6.2 Outlook
References
Internet of Vehicles: Security and Research Roadmap
1 Evolution of Automotive Connectivity
2 Autonomous Vehicle Standards
2.1 IEEE 1609
2.2 ISO/SAE DIS 21434 Automotive Cybersecurity Standard
3 Network Model
3.1 Various Connectivity Platforms
3.2 V2V Communication
3.3 V2I Communication
3.4 In-Vehicle Infrastructure
4 Security
4.1 CPS-IoT Design Challenges
4.2 Need for Security by Design
4.3 Design-Thinking Concepts
4.4 Cyberattacks in Recent Times
5 Security Objectives
6 Importance of Automotive Standards
7 Trust Management
7.1 Components of Trust
7.2 Attributes of Trust
7.3 Classification of Trust Management Models
8 Security Model for Automotive CPS
8.1 Storage Security
8.2 Computing Node Security
8.3 Communication Security
8.4 Sensor and Actuator Security
9 Privacy
9.1 Location Privacy
9.2 Data Privacy
10 Security for IoV
10.1 Threat, Vulnerability and Risk Analysis (TVRA)
10.2 Security Perimeter
10.3 Vehicular Attack Surfaces
11 Known Attacks on IoV Systems
11.1 Sensor/Actuator Attack
11.2 Remote Wireless Network Attacks
11.3 OTA Software Update Attacks
11.4 OBD-Based Attacks
12 Security Management
13 IoV Security Analysis: Research Roadmap
14 Security Analysis Through Simulation
15 Lightweight and Side-Channel Resistant Security Protocols
16 Post-Quantum Cryptography (PQC)
17 Conclusion
References
Part III Intrusion Detection Systems
Protecting Automotive Controller Area Network: A Review on Intrusion Detection Methods Using Machine Learning Algorithms
1 Introduction
1.1 Background and Motivation
1.2 Contributions and Outline
2 In-Vehicle Network Architecture
2.1 Evolution of In-Vehicle Network
2.2 The Necessity for Protecting CAN
3 Semantic-Based Intrusion Detection Methods
3.1 Motivation and Basic Idea
3.2 Machine Learning-Based Methods
3.3 Summary
4 Literal-Based Intrusion Detection Methods
4.1 Motivation and Basic Idea
4.2 Specification-Based Methods
4.3 Anomaly-Based Methods
4.4 Summary
5 Timing-Based Intrusion Detection Methods
5.1 Motivation and Basic Idea
5.2 Machine Learning-Based Methods
5.3 Summary
6 Signal Characteristics-Based Intrusion Detection Methods
6.1 Motivation and Basic Idea
6.2 Machine Learning-Based Methods
6.2.1 Signal Characteristics Derivation
6.2.2 Feature Extraction and Intrusion Detection
6.3 Summary
7 Conclusion
References
Real-Time Intrusion Detection in Automotive Cyber-Physical Systems with Recurrent Autoencoders
1 Introduction
2 Related Work
3 Background on Sequence Learning
3.1 Sequence Models
3.1.1 Recurrent Neural Networks (RNN)
3.1.2 Long Short-Term Memory (LSTM) Networks
3.1.3 Gated Recurrent Unit (GRU)
3.2 Autoencoders
4 Problem Definition
4.1 System Model
4.2 Communication Model
4.3 Attack Model
5 INDRA Framework Overview
5.1 Recurrent Autoencoder
5.1.1 Model Architecture
5.1.2 Training Process
5.2 Inference and Detection
6 Experiments
6.1 Experimental Setup
6.2 Intrusion Threshold Selection
6.3 Comparison of INDRA Variants
6.4 Comparison with Prior Works
6.5 IDS Overhead Analysis
6.6 Scalability Results
7 Conclusion
References
Stacked LSTM Based Anomaly Detection in Time-Critical Automotive Networks
1 Introduction
2 Related Work
2.1 Heuristi-Based Anomaly Detection
2.2 Machine Learning Based Anomaly Detection
3 Background
3.1 Recurrent Neural Network (RNN)
3.2 Long Short-Term Memory (LSTM) Network
3.3 Attention
4 Problem Formulation
4.1 System Overview
4.2 Communication Overview
4.3 Threat Model
5 Proposed Framework
5.1 Data Acquisition
5.2 Predictor Model
5.3 Detector Model
5.4 Model Testing
5.5 Anomaly Detection System Deployment
6 Experiments
6.1 Experimental Setup
6.2 Comparison of LATTE Variants
6.3 Comparison with Prior Works
6.4 Overhead Analysis
7 Conclusion
References
Deep AI for Anomaly Detection in Automotive Cyber-Physical Systems
1 Introduction
2 Related Work
3 TENET Framework: Overview
3.1 Data Collection and Preprocessing
3.2 Model Learning
3.3 Model Testing
3.3.1 Attack Model
3.3.2 Evaluation Phase
4 Experimental Setup
5 Conclusion
References
Physical Layer Intrusion Detection and Localization on CAN Bus
1 Introduction
2 Background
2.1 The CAN Protocol
2.2 Voltage-Based Intrusion Detection
2.3 Signal Propagation Delays
3 Localization Methods Based on Physical Layer Signals
3.1 Transmitter Identification by Propagation Delays
3.2 Signal Acquisition
3.3 The TIDAL-CAN Methodology
3.4 The CAN-SQUARE Methodology
4 Machine Learning on Physical Layer Signals
4.1 The ECUPrint Dataset
4.2 Results with Traditional Classifiers and Neural Networks
5 Discussion and Conclusion
References
Spatiotemporal Information Based Intrusion Detection Systems for In-Vehicle Networks
1 Introduction
2 In-Vehicle CAN Network
2.1 CAN Packets
2.2 Threat Model
3 Related Work
4 Weighted State Graph from CAN Frames
4.1 Constructing Weighted State Graph: Offline
4.2 Segmenting and Learning: Offline
4.3 Segmenting and Detecting: Online
4.4 Scoring the Subgraphs for Sliding Windows
4.5 Evaluation
4.5.1 Evaluation Metrics
4.5.2 Experiment Analysis
5 Spatiotemporal Information Based IDS for In-Vehicle Networks
5.1 Unknown Attack and Self-evolving Model
5.2 Spatiotemporal Information from CAN Frames
5.2.1 Convolutional LSTM Network
5.2.2 Robust and Self-evolving IDS for In-Vehicle Network
5.3 Evaluation
5.3.1 Threshold Selection for Classification
5.3.2 Experiment
5.3.3 Comparing the ConvLSTM Model with the LSTM Model
6 Conclusion
References
In-Vehicle ECU Identification and Intrusion Detection from Electrical Signaling
1 Introduction
2 System Model and Ringing Effect
2.1 Threat Models
2.2 Difference Between ECUs Voltage Outputs
2.3 Ringing Effect
2.3.1 From Dominant to Recessive States
2.3.2 From Recessive to Dominant States
3 Dominant States and Rising Edges for Source Identification
3.1 Signal Measurement and Preprocessing
3.2 Feature Extraction
3.3 Training and Testing
4 Evaluation
4.1 ECU Identification
4.1.1 Classification Algorithms
4.1.2 CAN Topology
4.1.3 CAN Signal States
4.1.4 On Real Vehicles
4.2 Intrusion Detection
4.2.1 Known ECUs
4.2.2 Unknown ECUs
4.3 Discussions
4.3.1 Environmental Factors
4.3.2 Sample Rate
4.3.3 Limitation and Battery/ECU Aging
5 Source Identification on In-Vehicle CAN-FD Networks
5.1 CAN-FD
5.2 System Model
5.3 Source Identification and Intrusion Detection
5.3.1 Experiment Setup
5.3.2 Sender Identification
5.3.3 Detecting Known ECUs
5.3.4 Detecting Unknown ECUs
5.4 Discussions
6 Conclusion
References
Machine Learning for Security Resiliency in Connected Vehicle Applications
1 Resiliency Needs and Challenges in CAV Applications
1.1 Constraints
1.2 ReDeM: Vision for ML-Based Resiliency
1.3 Overview of the Chapter
2 ReDeM Basics
2.1 Architecture
2.2 Appropriateness of ML-Based Solution
3 Architectural Considerations
3.1 Small Data Problem
3.2 Resource Constraints
3.3 Multi-Channel Adversary
3.4 Error Control and Recoverability
4 Design, Implementation, Tuning, and Validation of ML Component
4.1 Architecture Selection and Tuning
4.2 Data Preprocessing and Feature Selection
4.3 Decision Threshold Selection
4.4 Validation
5 ReDeM Case Study: Resilient Cooperative Adaptive Cruise Control
5.1 CACC Overview
5.2 Evaluation of ReDeM Resiliency on CACC
6 Conclusion
References
Part IV Robust Perception
Object Detection in Autonomous Cyber-Physical Vehicle Platforms: Status and Open Challenges
1 Introduction
2 Overview of Object Detectors
2.1 Two-Stage vs Single Stage Object Detectors
2.2 2D vs 3D Object Detectors
3 Deploying Object Detectors in AVs
3.1 Pruning
3.2 Quantization
3.3 Knowledge Distillation
4 Open Challenges and Opportunities
5 Conclusion
References
Scene-Graph Embedding for Robust Autonomous Vehicle Perception
1 Introduction
2 Scene-Graph Representation of Road Scenes
2.1 ADS Design Philosophies and Intermediate Representation
2.2 Graph-Based Driving Scene Understanding
2.3 Scene-Graph Extraction from Driving Scenes
3 Spatio-Temporal Scene-Graph Embedding Approach for Robust Automotive CPS Perception
3.1 Problem Formulation
3.2 Spatial Modeling
3.3 Temporal Modeling
3.4 Risk Inference
4 Experimental Results
4.1 Experimental Setup
4.2 Experiments on Risk Assessment
4.3 Evaluation of Attention Mechanisms on Risk Assessment
4.4 Transferability from Virtual To Real Driving
4.5 Risk Assessment By Action Type
5 Conclusion
References
Sensing Optimization in Automotive Platforms
1 Introduction
2 Related Work
3 Background
3.1 ADAS Features for Level 2 Autonomy
3.2 Feature Performance Metrics
4 VESPA Framework
4.1 Overview
4.2 Inputs
4.3 Design Space Exploration
4.3.1 SA + Greedy Random Adaptive Search Procedure (SA + Grasp)
4.3.2 Genetic Algorithm (GA)
4.3.3 Particle Swarm Optimization (PSO)
5 Experiments
5.1 Experimental Setup
5.2 Experimental Results
6 Conclusions
References
Unsupervised Random Forest Learning for Traffic Scenario Categorization
1 Introduction
2 Classification and Regression Trees
2.1 Computing the Optimal Split
2.2 Growing a Tree
3 Ensemble Learning with Random Forests
3.1 Ensemble Learning
3.2 Random Forests
3.2.1 Out-of-Bag Estimates
3.2.2 Proximity Measure
4 Random Forests for Unsupervised Learning
4.1 Similarity Measure Based on Random Forests
4.1.1 Constructing the Noise Dataset
4.1.2 Path Proximity
4.2 Hierarchical Clustering
4.3 Cluster Analysis and Visualization
5 Applications
5.1 Traffic Scenario Clustering
5.2 Open Set Recognition for Traffic Scenarios
6 Conclusion
References
Development of Computer Vision Models for Drivable Region Detection in Snow Occluded Lane Lines
1 Introduction
2 Methodology
2.1 Drive Cycles
2.2 Equipment and Instrumentation
2.2.1 Camera Sensor
2.2.2 Vehicle Type
2.3 Data Pipeline
2.3.1 Data Preparation
2.3.2 Image Ground Truth Labeling
2.3.3 Data Conditioning
2.4 Classical Machine Learning Models
2.4.1 Model Description
2.4.2 Model Training
2.5 Deep Neural Network Models
2.6 Model Architectures
2.6.1 Standard U-Net
2.6.2 Recurrent U-Net
2.6.3 Attention U-Net (Att U-Net)
2.6.4 Residual Operation
2.6.5 Residual + Attention U-Net (Res-Att U-Net)
2.6.6 Backbone U-Net
2.7 Model Training
2.8 Results
2.8.1 Classical Machine Learning Models
2.8.2 Convolutional Neural Network Models
2.8.3 Best ML Models vs Best CNN Model
2.9 Drivable Region Extraction from Tire Tracks
3 Conclusion
References
Machine Learning Based Perception Architecture Design for Semi-autonomous Vehicles
1 Introduction
2 Related Work
3 Background
3.1 ADAS Level 2 Autonomy Features
3.2 Sensor Placement and Orientation
3.3 Object Detection for Vehicle Environment Perception
3.4 Sensor Fusion
4 PASTA Architecture
4.1 Overview
4.2 Problem Formulation and Metrics
4.3 Design Space Encoder/Decoder
4.4 Design Space Exploration
4.4.1 Genetic Algorithm (GA)
4.4.2 Differential Evolution (DE)
4.4.3 Firefly Algorithm (FA)
4.5 Performance Evaluation
5 Experiments
5.1 Experimental Setup
5.2 Experimental Results
6 Conclusions
References
Part V Robust Control
Machine Learning and Optimization Techniques for Automotive Cyber-Physical Systems: Predictive Control During Acceleration Events to Improve Fuel Economy
Abbreviations
Abbreviations
1 Concept
1.1 Optimal EMS Mechanism
1.2 Model Details
1.2.1 Baseline EMS
1.2.2 Optimal EMS
1.3 Pre-Computing Optimal EMS for Approximate AE Prediction
1.4 Drive Cycle Simulations
2 Implementation
2.1 Baseline Torque Split Control
2.2 PAE Torque Split Control
3 Results
3.1 PAE Strategy Results
3.2 PAE vs Baseline Results
4 Conclusion
References
Learning-Based Social Coordination to Improve Safety and Robustness of Cooperative Autonomous Vehicles in Mixed Traffic
1 Introduction
2 Related Work
2.1 Multi-Agent Reinforcement Learning
2.2 Driver Behavior and Social Coordination
2.3 Safe and Robust Driving
3 Preliminaries and Formalism
3.1 Double Deep Q-Network
3.2 Driving Scenarios
3.3 Social Value Orientation for AVs
3.4 Autonomous Vehicles as Social Actors
3.5 Driving Behaviors
4 Problem Formulation
5 Safe and Robust Social Driving
5.1 Distinguishing Sympathy from Cooperation
5.2 Decentralized Social Reward
5.3 Deep MARL Architecture for Social Driving
5.4 Safety Prioritizer
5.5 Modeling Driver Behaviors
5.6 Implementation and Computational Details
6 Experiments and Results
6.1 Manipulated Variables
6.2 Performance Metrics
6.3 Hypotheses
6.4 Analysis and Results
6.4.1 Sensitivity Analyses
6.4.2 Domain Adaptation
6.4.3 Transfer Learning
6.4.4 Safety
6.4.5 Importance of Social Coordination
6.5 Qualitative Analyses
7 Concluding Remarks
References
Evaluation of Autonomous Vehicle Control Strategies Using Resilience Engineering
1 Introduction
2 Methodology
2.1 Simulation Environment
2.2 Development of Autonomous Driving Controllers
2.2.1 Pure Pursuit Controller
2.2.2 Deep Learning Controller
2.3 Resilience Engineering Applied to AVs
2.3.1 Resilience Engineering Overview
2.3.2 Operational Resilience Performance on AVs
2.4 Performance Evaluation Metrics
2.4.1 Traditional Metrics
2.4.2 Resilience Metrics
2.4.3 Standardization
3 Results
4 Conclusion
References
Safety-Assured Design and Adaptation of Connected and Autonomous Vehicles
1 Introduction
2 Safety Verification of Neural Network-Based Components in CAVs
2.1 Robustness Analysis of Deep Neural Networks
2.2 Safety Verification of Neural-Network Controlled Systems
3 System Adaptation and Design with Safety Assurance
3.1 Safety-Assured Runtime Adaptation
3.2 Safety-Driven Learning and System Design
4 Conclusion and Future Directions
References
Identifying and Assessing Research Gaps for Energy Efficient Control of Electrified Autonomous Vehicle Eco-Driving
1 Introduction
1.1 The Evolution of BEVs: The Modern Era
1.2 Energy Management and Energy Efficient Strategies for Electrified Vehicles
1.2.1 Powertrain EMS (P-EMS)
1.2.2 Eco-Routing (ER)
1.2.3 Eco-Driving (ED)
1.2.4 Summary
1.3 Automated Cyber-Physical Vehicles
2 Research Gap Derivation
2.1 AED System Architecture
2.2 Holistic Evaluation of System Maturity
2.2.1 Technology Readiness Levels (TRLs)
2.2.2 Integration Readiness Levels (IRLs)
2.2.3 System Readiness Levels (SRLs)
2.2.4 Research Gap Analysis Summary
3 Literature Review
3.1 Research Gap 1: Real-World AV Perception with Application to the AED Problem
3.2 Research Gap 2: Sparse or Missing Sensor Data on Global Derivation of AED
3.3 Research Gap 3: Performance of a Planning Subsystem Equipped with AED Integrated with a Physical Vehicle Plant
4 Conclusions
References
Index