Autonomous Systems in the Internet of Vehicles
Description
Advancements in sensor technology have enabled autonomous systems to operate efficiently and safely in the Internet of Vehicles environment. Multisensor image fusion is a crucial component in enhancing the capabilities of these autonomous systems by combining information from multiple sensors such as cameras, LiDAR, radar, and ultrasonic sensors. This book delves into the role of multisensor image fusion in the Internet of Vehicles for autonomous systems. It will cover the fundamental concepts of multisensor image fusion, different fusion methods, and their applications in autonomous systems for the IoV. It will also address the challenges associated with multisensor fusion, such as sensor calibration, synchronization, and noise reduction and discuss the benefits of multisensor fusion in improving object detection, tracking, and decision-making processes in autonomous vehicles operating in the IoV. This book is a comprehensive overview of multisensor image fusion in the context of IoV for autonomous systems, highlighting its importance in achieving reliable and robust autonomous navigation in dynamic and complex environments.
More details
Other editions
Additional editions
Persons
Balamurugan Balusamy, PhD is an Associate Dean at Shiv Nadar University with more than 12 years of experience. He has published more than 200 papers in international journals and edited and authored more than 80 books. His research focuses on engineering education, blockchain, and data sciences.
Sandeep Kumar Mathivanan, PhD is an Assistant Professor in the School of Computer Science and Engineering at Galgotias University with more than six years of research experience. He is a reviewer for a number of international journals and conferences. His research interests include machine learning, deep learning, remote sensing, and big data.
Prabhu Jayagopal, PhD is a Professor in the Department of Software and Systems Engineering in the School of Computer Science, Engineering, and Information Systems at the Vellore Institute of Technology. He has published 104 papers in international journals, book chapters, and conferences. His research interests include machine learning, artificial intelligence, and IoT related to healthcare.
S.K.B. Sangeetha, PhD is a Senior Assistant Professor in the Department of Computer Science and Engineering at the SRM Institute of Science and Technology with more than 15 years of teaching experience. She has published more than 75 research articles, ten book chapters in peer-reviewed international journals, and ten patents. She is a lifetime member of the International Society for Technology in Education and the International Education Initiative.
Ali Kashif Bashir, PhD is a Professor of Networks and Security at Manchester Metropolitan University. He is also affiliated with the University of Electronic Science and Technology of China, National University of Science and Technology, Pakistan, and University of Guelph. He has delivered more than 30 talks across the globe, organized more than 40 guest editorials, and chaired 35 conferences and workshops.
Content
Preface xi
1 A Cognitive Edge-Driven Autonomous Learning System for Scalable and Secure IoV Automation 1
V. Muthukumaran, S. Satheesh Kumar, Jahnavi S., Rose Bindu Joseph P. and Firoz Khan
1.1 Introduction 2
1.2 Related Study 3
1.3 System Methodology 7
1.3.1 Multilayer Edge Computing Framework 7
1.3.2 Federated Reinforcement Learning Model 10
1.3.3 Adaptive Dynamic Power Control Algorithm for CEALS 11
1.4 Experimentation Results 13
1.5 Conclusion 15
2 Adaptive Feature Alignment and Fusion for Multisensor Image Integration in the Internet of Vehicles 19
Vijay Anand R. and Madala Guru Brahmam
2.1 Introduction 20
2.2 Related Study 22
2.3 System Methodology 24
2.3.1 Multisensor Data Acquisition 24
2.3.2 Preprocessing 25
2.3.3 Dynamic Feature Alignment in AFAF-Net 25
2.3.4 Attention-Guided Fusion Method 26
2.3.5 Real-Time Object Detection 29
2.4 Experimentation Results 31
2.5 Conclusion 33
3 Design of ML-CASF: Multilayer Context-Aware Sensor Fusion for Autonomous Vehicles in the Internet of Vehicles 37
Sukumar R. and Sathishkumar V.E.
3.1 Introduction 38
3.2 Related Study 40
3.3 System Methodology 42
3.3.1 Sensor Data Acquisition 42
3.3.2 Preprocessing and Synchronization 42
3.3.3 Graph Construction for Sensor Data 42
3.4 Experimentation Results 48
3.5 Conclusion 52
4 Adaptive Multimodal Fusion for Robust Autonomous Driving Perception with Attention-Based Learning 55
Sangeetha R.
4.1 Introduction 56
4.2 Related Study 59
4.3 System Methodology 61
4.3.1 Data Collection and Preprocessing 61
4.3.2 Feature Extraction 62
4.3.3 Proposed Methodology 63
4.4 Experimentation Results 67
4.4.1 Performance Analysis 68
4.4.2 Computational Performance Comparison 69
4.4.3 Impact of Sensor Modalities on Detection Performance 70
4.5 Conclusion 71
5 Optimization-Driven Multisensor Fusion Framework for Autonomous Systems in the Internet of Vehicles 75
C. Gowdham, A.B. Hajira Be, C. Ashwini, S. Prabu and Zubair Rahaman
5.1 Introduction 76
5.2 Related Study 78
5.3 System Methodology 82
5.3.1 Data Acquisition and Preprocessing 82
5.3.2 Proposed Framework 83
5.3.2.1 EKF for Sensor Fusion 84
5.3.2.2 PF for Nonlinear Fusion 85
5.3.2.3 Deep Learning-Based Fusion Using CNNs and Transformers 85
5.4 Experimentation Results 86
5.5 Conclusion 89
6 A Hybrid Neurosymbolic Decision-Making Approach with Multimodal Sensor Fusion for Autonomous Vehicles 93
Devi A., Rose Bindu Joseph P. and Meram Munirathnam
6.1 Introduction 94
6.2 Related Study 96
6.3 System Methodology 100
6.3.1 Perception Module 100
6.3.2 Hybrid Decision-Making Algorithm for AVs 101
6.3.3 Trajectory Planning and Execution 103
6.4 Experimentation Results 103
6.5 Conclusion 105
7 Reinforcement Learning-Driven Multisensor Fusion for Real-Time Navigation in Intelligent and Opportunistic Vehicular Networks 109
Mahalakshmi, Suma T., Soya Mathew and Nitya S.
7.1 Introduction 110
7.2 Related Study 112
7.3 System Methodology 115
7.3.1 Perception Module 115
7.3.2 Proposed Algorithms 115
7.4 Experimentation Results 120
7.5 Conclusion 122
8 Hybrid Multimodal Fusion Network (HMFNet) for Enhanced Perception in Autonomous Vehicles 127
Mahalakshmi, Ranjini K. S., Nidhi S. Vaishnaw and Jesla Joseph
8.1 Introduction 128
8.2 Related Study 130
8.3 System Methodology 132
8.3.1 Dataset Used 132
8.3.2 Feature Extraction 133
8.3.3 Proposed HMFNet 134
8.4 Experimentation Results 138
8.5 Conclusion 140
9 Fusion-Enhanced Adaptive Learning for Robust Multisensor Integration in Autonomous IoV 143
A. Radha Krishna, U.V. Ramesh, S. Sathish Kumar and Aimin Li
9.1 Introduction 144
9.2 Related Study 148
9.3 System Methodology 151
9.3.1 Data Acquisition and Sensor Integration 151
9.3.2 SESW Algorithm 152
9.3.3 Multiscale Spatiotemporal Fusion Network 155
9.3.3.1 Feature Extraction Layer 155
9.3.3.2 Multiscale Fusion Module 155
9.3.3.3 Decision Refinement Layer 156
9.3.4 Multitask Output for Perception, Localization, and Path Planning 157
9.3.5 Final Computation Flow 157
9.4 Experimentation Results 158
9.4.1 Localization Accuracy in Simulation 159
9.4.2 Object Detection and Perception Accuracy 159
9.4.3 Computational Efficiency and Processing Latency 160
9.4.4 Decision-Making Latency with V2X Simulation 160
9.4.5 Path Planning and Collision Avoidance in Simulation 160
9.5 Conclusion 162
10 Dynamically Reconfigurable Multisensor Fusion for Enhanced Object Detection in Autonomous Vehicles 167
V. Muthukumaran, M. Sathish Kumar, G. Kumaran, Vidya K.B. and Ahmad Alkhayyat
10.1 Introduction 168
10.2 Related Study 170
10.3 System Methodology 173
10.3.1 Data Acquisition and Preprocessing 173
10.3.2 Proposed Algorithms 174
10.4 Experimentation Results 181
10.5 Conclusion 183
11 AI-Driven Edge Computing for Secure and Efficient Internet of Vehicles (IoV) Communication 187
Sukumar R. and Saurav Mallik
11.1 Introduction 188
11.2 Related Study 191
11.3 System Methodology 195
11.3.1 Data Collection and Preprocessing 195
11.3.2 Feature Extraction 197
11.3.3 Proposed Algorithms 197
11.4 Experimentation Results 201
11.5 Conclusion 207
12 Federated Autoencoder-GRU-Based Intrusion Detection System for Secure IoV-Connected Autonomous Vehicles 211
Pegadapelli Srinivas, Vijey Nathan, Radhika Rajavelu, Suresh Kulandaivelu and Roger Atanga
12.1 Introduction 212
12.2 Background Study on IoV 215
12.3 System Methodology 218
12.3.1 Dataset Description 218
12.3.2 Data Preprocessing 220
12.3.3 Proposed Federated Autoencoder-GRU IDS 221
12.4 Experimental Results 225
12.5 Conclusion 229
13 Edge-Driven Multimodal Fusion Framework for Real-Time Emotion-Aware Vehicular Networks 233
Manjula Sanjay Koti, S. Satheesh Kumar, Janani S., Arun A. and Mahmoud Ahmad Al-Khasawneh
13.1 Introduction 234
13.2 Related Study 238
13.3 System Methodology 243
13.3.1 Multimodal Data Acquisition 243
13.3.2 Signal Preprocessing and Synchronization 245
13.3.3 Feature Extraction and Fusion 246
13.3.4 Emotion Recognition Engine 248
13.3.5 Emotional Readiness for Control Handover 250
13.4 Experimentation Results 253
13.5 Conclusion 257
14 Spatiotemporal Attention-Based CNN-BiLSTM Model for Robust Lane and Obstacle Detection in IoV-Enabled Autonomous Driving 261
Suresh Kulandaivelu, Syied Mazar, Sangeetha N., Sathiyapriya Rajavelu and Anita Garhwal
14.1 Introduction 262
14.2 Related Study 265
14.3 System Methodology 269
14.3.1 Dataset Used and Preprocessing 269
14.3.2 Network Architecture: Spatiotemporal Attention-Enhanced CNN-BiLSTM 272
14.3.3 Inference Optimization and Real-Time Deployment 274
14.4 Experimentation Results 275
14.5 Conclusion 279
15 Multimodal Vision-LiDAR Transformer Fusion for End-to-End IoV-Based Autonomous Navigation 283
Mohan Mani, Hariprasath K., C. Vijayakumar, Sathiyapriya Rajavelu and Sarawoot Boonkirdram
15.1 Introduction 284
15.2 Background Study 287
15.3 System Methodology 290
15.3.1 Simulation Environment and Dataset Generation 290
15.3.2 Multimodal Preprocessing Pipeline 291
15.3.3 Network Architecture: Transformer-Based Multimodal Fusion 293
15.4 Experimental Results 298
15.5 Conclusion 302
References 303
Index 305
1
A Cognitive Edge-Driven Autonomous Learning System for Scalable and Secure IoV Automation
V. Muthukumaran1*, S. Satheesh Kumar2, Jahnavi S.3, Rose Bindu Joseph P.4 and Firoz Khan5
1Department of Mathematics, College of Engineering and Technology, SRM Institute of Science and Technology, Kattankulathur, India
2School of Information Science, Presidency University, Bangalore, Karnataka, India
3Department of AIML, B.M.S. College of Engineering, Bangalore, India
4Department of Mathematics, Dayananda Sagar College of Engineering, Bangalore, India
5Faculty in Center for Information and Communication Sciences, Ball State University, Muncie, IN., U.S.A.
Abstract
The rapid evolution of the Internet of Vehicles (IoV) has led to a transformative shift in various industries, enabling seamless connectivity, real-time monitoring, and intelligent automation. The study proposes a novel cognitive edge-driven autonomous learning system (CEALS), which maximizes IoV-based automation through the integration of edge artificial intelligence, federated reinforcement learning, and blockchain-based security. CEALS's federated reinforcement learning model improves adaptive decision-making precision by 41.3% compared to conventional deep learning-based IoV automation, and its multilayer edge computing architecture reduces latency by 52.7%. With an attack detection rate of 97.8%, a blockchain-embedded security mechanism ensures data integrity and improves on existing security measures by 18.6%. Dynamic power control algorithms reduce energy consumption by 38.9% and improve overall system effectiveness by 45.2% based on experimental verification in an industrial IoV large-scale network. The proposed CEALS architecture is a scalable, highly efficient next-generation smart environment solution, pushing the limits of autonomous IoV networks.
Keywords: Cognitive edge computing, autonomous learning system, Internet of Vehicles, federated reinforcement learning, blockchain security, latency reduction, energy efficiency, smart environment
1.1 Introduction
The industrial automation era was energized by programmable logic controllers; autonomous systems began in the mid-20th century with mechanical control systems. Hand changes were necessary because the systems were fixed and inflexible [1]. Rule-based systems dominated with the expansion of computing power, so there was some allowance for decision-making. But in dynamic and uncertain environments, these were challenging, so more advanced intelligence needed to be integrated [2]. The advent of the Internet of Vehicles (IoV) in the early 2000s transformed autonomous systems by making it possible for real-time sharing of data between networked objects. Although cloud computing was beneficial for data processing, it also brought with it central dependencies, bandwidth limitations, and latency, which hindered real-time autonomous decision-making in mission-critical applications such as smart vehicles [3].
Various methods of enhancing IoV-based autonomous systems have been researched to cope with these concerns. Traditional rule-based automation is less pliable to handle dynamic conditions [4]. Decision-making powered by machine learning (ML) is also chosen due to cloud computing along with issues of data privacy along with ultrahigh latency. The solution came in the rise of edge computing reducing latency by locating processing nearer to IoV devices. But where complex situations necessitated coordination among sets of nodes in large numbers, the absence of collective information within isolated edge devices saw bad decision-making [5]. Federated learning (FL), which had provided decentralized learning for IoV devices with a view to mitigating privacy concerns, was beset with high communication overhead and slow convergence rates. Security attacks were also common, such as network intrusions, malware attacks, and data tampering [6].
The limitations of the solutions that are currently available represent the demand for a better, scalable, and secure solution [7, 8]. Creating a smart autonomous system that enhances real-time decision-making, addresses latency, enhances security, and maximizes the utilization of resources in huge-scale IoV networks is the primary driving force of this work. With growing IoV networks in diverse applications such as predictive maintenance and intelligent transport, autonomous systems with reduced or no human intervention are increasingly being sought after. Effective coordination in decentralized IoV nodes with ensured security, scalability, and power efficiency is the primary concern. In order to create an IoV system capable of making highly adaptive, reliable decisions in real time, one needs an innovative solution with FL, edge intelligence, and blockchain security capabilities.
The study introduces the cognitive edge-driven autonomous learning system (CEALS), a new generation of autonomous IoV system architecture with multilayer edge computing, federated reinforcement learning, and blockchain-based security to solve the issues as mentioned above. For real-time processing of information and decision-making, CEALS utilizes a hierarchical edge computing model. CEALS facilitates hierarchical coordination among edge layers to achieve maximum learning efficiency over traditional edge computing systems through decentralized decision-making at nodes. The system also ensures persistent optimization of decentralized decision-making and relieves communication overhead through federated reinforcement learning. CEALS is very secure and reliable for mass deployment as it integrates blockchain technology, further protecting data exchange by decentralizing trust, rendering the data tamper-proof, and minimizing cybersecurity attacks.
The main contributions of the study is to develop a CEALS that integrates edge artificial intelligence (AI), federated reinforcement learning, and blockchain-based security to enhance real-time decision-making, security, and energy efficiency in large-scale IoV networks.
1.2 Related Study
The complexity of networked systems has created immense interest in developing secure and autonomous IoV systems. Blockchain-based ASSERT architecture is targeted at improving security in self-adaptive Internet of Things (IoT) systems [9]. The method uses decentralized trust models and smart contracts to provide secure and resilient IoT ecosystems. ASSERT's application of blockchain technology enables adaptive and dynamic system tuning as well as minimizes typical security weaknesses such as data tampering and unauthorized access. Improvement in security assurance is demonstrated in the research to be concentrating on the area of how the application of blockchain makes self-adaptive IoT infrastructure more reliable for real-time use.
For massive-scale IoT deployment in multidomain IoT systems, Secure Interconnected Autonomous System Architecture (SIASA) has been proposed [10]. The architecture provides a security solution in the form of layered architecture with intrusion detection, cryptographic primitives, and ML-based anomaly detection to support secure IoT communication in heterogeneous systems. The study shows how much secure data exchange and cross-domain compatibility are required for massive-scale IoT deployments. Based on performance ratings, SIASA is an excellent solution to scalable and secure IoT networks deployed to a broad range of dynamic scenarios because it reduces security events and improves self-healing system robustness.
Self-validation in real-time for security in situations is managed by an industrial IoT context-aware security assertion infrastructure [11]. An AI-based context-variable adaptive security model has been discussed in this study. False alarms are kept to a minimum and detection made precise with the help of ML-based anomaly detection, which adjusts security policies dynamically in terms of the condition of operations. The results indicate that the security processes with AI can be implemented in autonomous industrial IoT networks to enhance reliability and resistance against cyber-attacks, depending on the high accuracy of the security statement.
The main motivation for an analysis of IoT-based cloud applications' autoscaling mechanisms is system resource utilization and scalability maximization [12]. Comparison of advantages and disadvantages of different autoscaling systems utilized so far is done on the basis of their classification as rule-based, predictive, and reinforcement learning-based systems. The study concludes that, although rule-based autoscaling is reactive in nature, it can never be responsive enough to address workload variations. Although improved, prediction and reinforcement learning-based methods are computationally intensive. In future IoT-cloud applications, the research suggests the necessity of hybrid autoscaling models with tradeoffs between responsiveness, scalability, and energy efficiency.
For facilitating collective, individual, and real-time learning of cyber-physical systems, an IoT three-layer architecture is envisioned [13]. A hierarchical processing paradigm is designed for this architecture, which distributes the computation among edge, fog, and cloud layers optimally. The approach improves real-time decision-making, decreases latency, and scales better in IoT applications with continuous learning requirements. The study illustrates how strict three-layer architecture restricts communication overhead and optimizes resource utilization to improve the...
