Deep Learning and Its Applications for Vehicle Networks

Description

Deep Learning (DL) is an effective approach for AI-based vehicular networks and can deliver a powerful set of tools for such vehicular network dynamics. In various domains of vehicular networks, DL can be used for learning-based channel estimation, traffic flow prediction, vehicle trajectory prediction, location-prediction-based scheduling and routing, intelligent network congestion control mechanism, smart load balancing and vertical handoff control, intelligent network security strategies, virtual smart and efficient resource allocation and intelligent distributed resource allocation methods. This book is based on the work from world-famous experts on the application of DL for vehicle networks. It consists of the following five parts: (I) DL for vehicle safety and security: This part covers the use of DL algorithms for vehicle safety or security. (II) DL for effective vehicle communications: Vehicle networks consist of vehicle-to-vehicle and vehicle-to-roadside communications. This part covers how Intelligent vehicle networks require a flexible selection of the best path across all vehicles, adaptive sending rate control based on bandwidth availability and timely data downloads from a roadside base-station. (III) DL for vehicle control: The myriad operations that require intelligent control for each individual vehicle are discussed in this part. This also includes emission control, which is based on the road traffic situation, the charging pile load is predicted through DL andvehicle speed adjustments based on the camera-captured image analysis. (IV) DL for information management: This part covers some intelligent information collection and understanding. We can use DL for energy-saving vehicle trajectory control based on the road traffic situation and given destination information; we can also natural language processing based on DL algorithm for automatic internet of things (IoT) search during driving. (V) Other applications. This part introduces the use of DL models for other vehicle controls. Autonomous vehicles are becoming more and more popular in society. The DL and its variants will play greater roles in cognitive vehicle communications and control. Other machine learning models such as deep reinforcement learning will also facilitate intelligent vehicle behavior understanding and adjustment. This book will become a valuable reference to your understanding of this critical field. Cover Half Title Title Page Copyright Page Table of Contents Preface About the Editors List of Contributors Part I: Deep learning for vehicle safety and security Chapter 1: Deep learning for vehicle safety 1.1 Introduction 1.2 Deep learning for internal vehicle monitoring 1.2.1 Camera-based system 1.2.2 Wearable sensor-based system 1.2.3 Driver behavior monitoring 1.3 Deep learning for surrounding environment perception 1.3.1 Road detection 1.3.2 Vehicle surrounding environment detection 1.3.3 Object detection in challenging environments 1.4 Deep learning for traffic management 1.4.1 Traffic flow modelling 1.4.2 Vehicle and infrastructure communications 1.5 Deep learning-based route planning and navigation 1.5.1 Route planning for travellers 1.5.2 Route planning for food transportation 1.5.3 Dynamic routing with unknown map 1.6 Conclusions References Chapter 2: Deep learning for driver drowsiness classification for a safe vehicle application 2.1 Introduction 2.1.1 Importance of drowsiness detection 2.1.2 Application in future automated vehicles 2.2 Driver drowsiness detection methods 2.2.1 Subjective measures 2.2.2 Objective measures 2.2.2.1 Input sources for driver drowsiness detection 2.2.3 Deep learning methods 2.2.3.1 Deep learning methods applied to the biosignals 2.3 Comparison of methods 2.4 Summary and outlook Notes References Chapter 3: A deep learning perspective on Connected Automated Vehicle (CAV) cybersecurity and threat intelligence 3.1 Introduction 3.2 CAV technological enablers: automation and connectivity 3.3 CAV threat landscape and threat intelligence 3.3.1 In-vehicle (low-level sensor) cyber vulnerabilities 3.3.2 Vehicle control modules 3.3.3 Security analysis of CAV threats 3.3.4 Attack surfaces 3.3.5 Organizational risks to CAV ecosystem 3.4 CAV threat mitigation: anomaly detection and classification with deep learning 3.5 Frontiers in deep learning (advancement and future) 3.5.1 Meta-learning 3.5.2 Federated learning 3.6 End-to-end deep CNN-LSTM architecture for CAV cyberattack detection 3.6.1 Performance analysis 3.6.1.1 Dataset 3.6.1.2 Evaluation metrics 3.6.2 Results and discussions 3.7 Conclusion References Part II: Deep learning for vehicle communications Chapter 4: Deep learning for UAV network optimization 4.1 Introduction 4.2 Key categories for UAV networking throughput enhancement 4.3 Routing enhancement for UAV networking throughput 4.3.1 Position-based routing 4.3.1.1 Single path-based 4.3.1.2 Multiple path-based 4.3.2 Topology-based routing 4.3.2.1 Proactive 4.3.2.2 Reactive 4.3.2.3 Hybrid 4.3.3 Swarm-based routing 4.3.4 DL-enabled routing for UAV networking 4.4 UAV networking construction 4.4.1 UAV swarm networking construction 4.4.2 DL-enabled UAV swarm networking enhancement 4.5 DL-enabled UAV networking throughput 4.5.1 DL-enabled allocation for throughput enhancement 4.5.2 DL-enabled scheduling for throughput enhancement 4.5.2.1 Scheduling for UAV networking 4.5.2.2 DL-enabled scheduling for UAV networking 4.6 Conclusions References Chapter 5: State-of-the-art in PHY layer deep learning for future wireless communication systems and networks 5.1 Introduction 5.1.1 Related survey papers 5.1.2 Summary of this chapter 5.2 Data-driven ML methods for transceiver optimization 5.2.1 Data-driven approach for end-to-end transceiver optimization 5.2.2 Model-aided data-driven methods for modular transceiver optimization 5.3 Deep learning for symbol detection 5.3.1 Incorporating expert knowledge into autoencoders 5.3.2 Implementing NNs at the receiver 5.3.3 Sequential detectors using ML 5.4 Channel estimation using ML 5.5 Channel prediction in frequency- and time-domain using ML 5.6 Channel coding using AI/ML 5.7 Intelligent link adaptation 5.8 Intelligent radios 5.8.1 Intelligent spectrum sensing 5.8.2 Automatic signal recognition using CNNs 5.8.3 Intelligent radio environment 5.9 ML for system-level performance evaluation of wireless networks 5.10 Conclusions Notes References Chapter 6: Deep learning-based index modulation systems for vehicle communications 6.1 Introduction 6.2 V2V and V2I Communications 6.3 Deep learning-based index modulation systems 6.3.1 Multicarrier-based index modulation systems 6.3.1.1 Transmitter of system model with OFDM-IM 6.3.1.2 Traditional detection scheme 6.3.1.3 Deep learning-based detector 6.3.1.3.1 Deep learning model 6.3.1.3.2 Training procedure 6.3.1.3.3 Online deployment 6.3.2 Single-carrier based index modulation systems 6.3.2.1 Transmitter of system model with CIM-SS 6.3.2.2 Conventional detection scheme 6.3.2.3 Deep learning-based detection 6.3.3 Multi-input multi-output based index modulation systems 6.3.3.1 System model and related works 6.3.3.2 Deep learning-based SMTs 6.3.3.3 Performance analysis on energy efficiency of vehicle communications with IM 6.4 Conclusions References Chapter 7: Deep reinforcement learning applications in connected-automated transportation systems 7.1 Introduction 7.1.1 Chapter organization 7.2 Deep reinforcement learning: theory and background 7.2.1 (Deep) reinforcement learning: a brief history 7.2.2 Classical reinforcement learning 7.2.2.1 Value-based RL 7.2.2.2 Policy-based RL 7.2.3 Deep reinforcement learning 7.2.3.1 Deep Q-networks 7.2.3.2 Deep policy networks 7.2.3.3 Deep actor-critic networks 7.2.4 Formulating (deep) reinforcement learning for CAV application 7.3 Data environment in CAV networks 7.3.1 Benefits 7.3.2 Data generated by AVs 7.4 Deep reinforcement learning applications: connected vehicles 7.4.1 Lane changing and assistance 7.4.2 Traffic signal control 7.4.3 Traffic flow optimization 7.4.4 Rail and maritime transportation 7.4.5 Data communications, computing, and networking 7.4.6 DRL applications for cybersecurity 7.5 Deep reinforcement learning applications: automated driving systems 7.5.1 Motion planning 7.5.2 Lateral control 7.5.3 Safety 7.6 Challenges and future directions 7.6.1 Transferability to real-world applications 7.6.2 Representation of traffic environment 7.6.3 Formulating reward functions 7.6.4 Multi-agent DRL in CAV environment 7.6.5 Partial state observability References Part III: Deep learning for vehicle control Chapter 8: Vehicle emission control on road with temporal traffic information using deep reinforcement learning 8.1 Introduction 8.2 Related work 8.3 Overview 8.3.1 Preliminary 8.3.2 Traffic data insight 8.3.3 Problem formulation 8.4 Methodology 8.4.1 Framework 8.4.2 EFRL model 8.5 Experiments 8.5.1 Data and setup 8.5.2 Baselines and metric 8.5.3 Results 8.6 Conclusion References Chapter 9: Load prediction of an electric vehicle charging pile 9.1 Introduction 9.2 Charging load characteristic analysis of electric vehicles 9.3 Quantile regression model of dilated causal convolutional 9.3.1 Dilated causal convolutional 9.3.2 Kernel density estimation 9.3.3 Dilated causal convolutional quantile regression 9.3.4 Model evaluation index 9.3.5 Example simulation based on python 9.4 Spatio-temporal dynamic load prediction of charging pile load based on deep learning 9.4.1 Spatio-temporal dynamic load prediction of the charging pile 9.4.2 Spatio-temporal dynamic load matrix construction 9.4.3 Spatio-temporal convolutional networks model 9.4.4 Spatio-temporal dynamic load forecasting based on dilated causal convolution-2D 9.4.5 Spatio-temporal dynamic load forecasting based on Spatio-temporal neural network 9.4.6 Example simulation based on python 9.5 Conclusions References Chapter 10: Deep learning for autonomous vehicles: A vision-based approach to self-adapted robust control 10.1 Introduction 10.2 References selection via deep learning image processing 10.2.1 CNN analytic outcomes as control references 10.2.2 Experimental data 10.2.3 Multi-objective evaluation 10.2.4 Control state variable 10.3 Robust control design 10.3.1 System identification 10.3.2 Robust Linear Quadratic Regulator (RLQR) 10.3.3 H∞ Controller 10.4 Case study for hybrid controller 10.4.1 Simulation environment and problem objective 10.4.2 Machine learning design 10.4.2.1 Input and output 10.4.2.2 Reward function 10.4.2.3 Neural network design 10.4.2.4 Simulation technical details 10.4.3 Hybrid control design 10.4.3.1 Control reference selection 10.4.3.2 System identification and nominal LQR 10.4.3.3 Uncertainty estimation via evolutionary search 10.4.4 Performance evaluation 10.5 Conclusions Note References Part IV: DL for information management Chapter 11: A natural language processing-based approach for automating IoT search 11.1 Introduction 11.2 IoT search engine 11.2.1 Architecture 11.2.2 Key components 11.2.3 Research challenges 11.3 NLP-based query processing 11.3.1 Design rationale 11.3.2 Basic components of NLP 11.3.3 NLP tools 11.3.4 Comparison of NLTK and spaCy 11.4 The ACQUISE approach 11.4.1 Baseline strategy 11.4.2 Enhanced static strategy 11.4.3 Enhanced dynamic strategy 11.5 Performance evaluation 11.5.1 Methodology 11.5.2 Results 11.6 Discussion 11.6.1 Machine learning 11.6.2 Protocols and algorithms 11.6.3 Security and privacy 11.7 Related work 11.8 Final remarks Acknowledgement References Chapter 12: Toward incentive-compatible vehicular crowdsensing: A reinforcement learning-based approach 12.1 Introduction 12.2 Edge-assisted vehicular crowdsensing 12.2.1 Architecture design 12.2.2 Workflow 12.3 Incentive mechanism for vehicle recruitment 12.3.1 Stackelberg game 12.3.2 Strategy of the SSP 12.3.3 Strategies of vehicles 12.4 Case study 12.5 Conclusions Appendix A References Chapter 13: Sub-signal detection from noisy complex signals using deep learning and mathematical morphology 13.1 Introduction 13.2 LSTM-RNN and mathematical morphology-based algorithm to detect sub-signals from noisy complex signals 13.2.1 Data preparation and pre-processing 13.2.2 LSTM-RNN local sub-signal learning 13.2.3 Mathematical morphological global sub-signal detection 13.3 Experimental results 13.4 Conclusion Acknowledgment Notes References Part V: Miscellaneous Chapter 14: The basics of deep learning algorithms and their effect on driving behavior and vehicle communications 14.1 Introduction 14.2 Basics of deep learning algorithms and supervised learning 14.2.1 Linear regression and logistic regression 14.2.2 Artificial Neural Networks 14.2.3 Convolutional Neural Networks 14.2.4 Recurrent Neural Networks 14.2.5 Deep learning architectures 14.3 Deep unsupervised and semi-supervised learning 14.3.1 Restricted Boltzmann Machines and deep belief nets 14.3.2 Autoencoders & variational autoencoders 14.3.3 Generative adversarial networks 14.3.4 Transformers 14.4 Hyperparameters, pre-processing and optimization 14.4.1 Data augmentation and transfer learning 14.4.2 Weight initialization, activation functions and optimizers 14.4.3 Training time, pre-processing and architectural refinements 14.5 Applications of deep learning in driving behavior analysis and vehicle communication 14.6 Conclusions References Chapter 15: Integrated simulation of deep learning, computer vision and physical layer of UAV and ground vehicle networks 15.1 Introduction 15.2 Applications that can benefit from CAVIAR simulations 15.2.1 Simulation of UAV-enabled AI/ML 15.2.2 Beam-selection for V2I 15.3 Multi-domain integrated simulators 15.3.1 Wireless channel generation with Raymobtime 15.3.2 Caviar simulations 15.4 Simulations results 15.4.1 Beam selection for V2I with lidar as input 15.4.2 In-loop CAVIAR simulation of a computer vision application 15.4.3 Impact of 3D model accuracy on wireless channels 15.5 Conclusions Acknowledgments Notes References