Artificial Intelligence for Future Networks

Mohammad A. Matin is a Professor and Chairman in the Department of Electrical and Computer Engineering at North South University, Dhaka, Bangladesh. Sotirios K. Goudos is a Professor in the Department of Physics at the Aristotle University of Thessaloniki, Greece and the Director of the ELEDIA@AUTH lab member of the ELEDIA Research Center Network. George K. Karagiannidis is a Professor in the Department of Electrical and Computer Engineering of Aristotle University of Thessaloniki, Greece, and the Head of the Wireless Communications and Information Processing (WCIP) Group.

• About the Editors xvList of Contributors xviiAcknowledgments xxi1 Intelligent Beam Prediction and Tracking 1Christos Masouros, Jianjun Zhang, and Yongming Huang1.1 Introduction 11.2 Challenge of Beam Prediction Modeling in Wireless Communications 51.3 Prior Identification – Perspective of Function Space 71.3.1 Perspective of Function Space 81.3.2 Useful Priors for Beam Process Modeling 91.3.2.1 High-speed Train Communication 91.3.2.2 Indoor Environment 91.3.2.3 City Street Environment 91.4 Methodology from Stochastic Process 121.5 Stochastic Continuity – Beam Index Difference 161.5.1 Beam Index Difference Technique 161.5.2 BPT Solution via Beam Index Difference 171.5.3 Theoretical Analysis for Beam Index Difference 211.6 Stochastic Smoothness – Hybrid Data-induced Kalman Filtering 251.6.1 Theoretical Foundation 261.6.2 Implicit Dynamics Learning via Multitask Learning 281.6.3 SDE Representation and Efficient Inference 311.7 Beam Width Optimization 331.7.1 Stochastic Continuity – Locality Principle of Beam Change and Data Transmission with Multiresolution Beam 331.7.2 Stochastic Smoothness – Low-frequency Sounding via BWO and Long-term Prediction 351.8 Numerical Results 361.8.1 Simulation Results for Stochastic Continuity 371.8.2 Simulation Results for Stochastic Smoothness 391.9 Conclusion 45References 462 Signal Detection with Machine Learning 51Jayakrishnan Vijayamohanan, Arjun Gupta, Manel Martínez-Ramón, and Christos Christodoulou2.1 Introduction 512.2 Symbol Detection 522.2.1 The Viterbi Algorithm 522.2.2 Channel Equalization Through Machine Learning 542.2.3 Machine Learning Implementations of the Viterbi Algorithm 572.3 Modulation Detection 602.3.1 Signal Model 612.3.2 Feature Selection 622.3.3 Maximum Likelihood Estimation 642.3.4 Neural Modulation Detection 642.3.4.1 Convolutional Neural Network 652.3.4.2 CNN Modulation Detection 672.4 Source Detection 742.4.1 Array Signal Model 742.4.2 Conventional Source Detection 772.4.3 Neural Source Detection 792.4.3.1 CNN Detector 802.4.3.2 RadioNet 822.5 Conclusion 84References 853 AI-Aided Channel Prediction 93Oscar Stenhammar, Gábor Fodor, and Carlo FischioneAcronyms 933.1 Introduction 943.1.1 Channel Aging 943.1.2 Channel Estimation 963.1.3 Channel Prediction 963.2 Preliminaries 983.2.1 Multilayer Perceptron 983.2.2 Convolutional Neural Network 1003.2.3 Recurrent Neural Network 1013.2.3.1 Long Short-Term Memory 1013.2.3.2 Gated Recurrent Units 1033.2.4 Transformer 1033.3 Previous Work 1053.3.1 Previous Work in Channel Estimation 1053.3.2 Conventional Channel Prediction 1073.3.3 Previous Work in AI-Aided Channel Prediction 1093.4 Experimental Evaluations 1133.4.1 Simulation Setup 1133.4.2 Neural Network Setup 1153.4.3 Experimental Results 1183.5 Discussion 1213.6 Summary 123References 1244 Semantic Communications 131Qiyang Zhao, Hang Zou, Mehdi Bennis, and Merouane Debbah4.1 Introduction 1314.2 Semantic Information and Semantic-Native Communication 1344.2.1 Semantic Information Theory 1344.2.2 Semantic-Native Communication 1374.3 Interplay of AI and Semantic Communication 1404.3.1 AI for Semantic Communication 1404.3.2 Semantic-Native Collective Intelligence 1434.4 Conclusion 145References 1465 Federated Learning for Wireless Communications 151Ahmet M. Elbir and Wei Shi5.1 Introduction 1515.2 Channel Models 1555.2.1 mmWave Channel Model 1555.2.2 THz Channel Model 1575.2.2.1 Near-Field Array Model 1585.2.2.2 Near-Field Beam Squint 1605.3 Federated Learning for Channel Estimation 1625.3.1 Training Data Collection 1625.3.2 FL-Based Model Training 1635.3.3 FL for mmWave Channel Estimation in Massive MIMO 1655.3.4 FL for mmWave Channel Estimation in RIS-Assisted Massive Mimo 1695.3.5 FL for THz Channel Estimation 1725.4 FL For Hybrid Beamforming 1765.5 Conclusions 178Acknowledgment 179References 1796 Federated Learning in Mesh Networks 185Xu Wang, Yuanzhu Chen, and Octavia A. Dobre6.1 Introduction 1856.1.1 Federated Learning 1856.1.2 Mesh Networks 1866.1.3 The Convergence: Federated Learning on Mesh Networks 1876.2 Decentralized Federated Learning 1886.2.1 Traditional Federated Learning versus Decentralized Federated Learning 1896.2.2 Core Principles of Decentralized Federated Learning 1916.2.3 Advantages of Decentralization in Federated Learning 1916.2.4 Architecture Variants for Decentralized Federated Learning 1926.2.5 Challenges of Decentralization in Federated Learning 1926.3 Mesh Networks 1926.3.1 Why Mesh Networks 1936.3.2 Fundamental Concepts and Terminologies 1936.3.3 Topological Structures 1936.3.4 Advantages of Mesh Networks 1946.3.5 Challenges and Limitations 1956.3.6 Integration with Federated Learning 1956.4 The Intersection: Decentralized Federated Learning over Mesh Networks 1966.4.1 Natural Synergy Between Federated Learning and Mesh Networks 1966.4.2 Potential Benefits of the Convergence 1966.4.3 Enabling Technologies 1986.4.4 Challenges at the Intersection 1986.4.4.1 Communication Overhead 1986.4.4.2 Data Heterogeneity and Non-IID Data 1996.4.4.3 Model Aggregation in Decentralized Networks 1996.4.4.4 Network Latency and Asynchrony 1996.4.4.5 Security and Privacy Concerns 1996.4.4.6 Scalability Concerns 2006.4.4.7 Fault Tolerance and Robustness 2006.4.4.8 Resource Constraints 2006.5 Solutions 2006.5.1 Communication Overhead 2006.5.2 Data Heterogeneity and Non-IID Data 2016.5.3 Model Aggregation in Decentralized Networks 2016.5.4 Latency and Asynchrony 2026.5.5 Security and Privacy Concerns 2026.5.6 Scalability Concerns 2026.5.7 Fault Tolerance and Robustness 2036.5.8 Resource Constraints 2036.6 State-of-the-Art and Noteworthy Implementations 2046.6.1 Decentralized Federated Learning Techniques 2046.6.1.1 Network Topology 2046.6.1.2 Communication Protocols 2046.6.1.3 Privacy Enhancements 2056.6.2 Advances in Mesh Networking Technologies 2056.6.2.1 Low-Latency Protocols 2056.6.2.2 Scalable Architectures 2066.6.2.3 Security Enhancements 2066.6.3 Decentralized Federated Learning on Mesh Networks: Integrated Approaches 2066.6.4 Toolkits and Platforms 2076.6.5 Benchmarks and Evaluation 2086.7 Future Directions and Open Research Challenges 2096.7.1 Advanced Algorithms 2096.7.2 Enhanced Security Mechanisms 2096.7.3 Network Optimization 2106.7.4 Interoperability and Standardization 2106.7.5 Energy Efficiency and Sustainability 2116.7.6 User-Centric Approaches 2116.7.7 Real-time Decentralized Federated Learning 2126.7.8 Codesigning Hardware and Software 2126.7.9 Ethical and Regulatory Considerations 2136.7.10 Interdisciplinary Research 2136.8 Concluding Remarks 213References 2147 Antenna Design Using Artificial Intelligence 227Sotirios K. Goudos, Mohammad A. Matin, and George K. Karagiannidis7.1 Introduction 2277.2 Evolutionary Algorithms 2297.2.1 Mainstream Algorithms 2297.2.1.1 Genetic Algorithms 2297.2.1.2 Particle Swarm Optimization 2307.2.1.3 Differential Evolution 2317.2.1.4 Ant Colony Optimization 2327.2.2 Emerging Algorithms 2357.2.2.1 Biogeography-Based Optimization 2357.2.2.2 Grey Wolf Optimizer 2357.2.2.3 Wind-Driven Optimization 2357.2.2.4 Salp Swarm Algorithm 2357.2.2.5 Artificial Bee Colony (ABC) 2367.2.2.6 Harmony Search (HS) 2367.2.2.7 Shuffled Frog-Leaping Algorithm 2377.2.3 Antenna Optimization Using Evolutionary Algorithms 2377.2.3.1 Problem Formulation 2377.2.3.2 Numerical Results 2397.3 Machine Learning 2447.3.1 Artificial Neural Networks (ANNs) 2447.3.2 Support Vector Machines 2447.3.3 Gaussian Process (GP) 2457.3.4 Deep Learning (DL) 2457.3.5 ANFIS 2457.3.6 Surrogate Modeling 2467.3.6.1 Surrogate Modeling Example 2487.4 Knowledge Representation 2527.5 Conclusion 253References 2538 AI-Driven Approaches for Solving Electromagnetic Inverse Problems 257Marco Salucci, Maokun Li, and Andrea Massa8.1 Introduction 2578.2 Mathematical Formulation 2588.3 AI-Based EM–IP Solution Strategies 2628.3.1 3-Step Learning-by-Examples (LBE) Framework 2638.3.2 System-by-Design (SbD) Framework 2678.3.3 Deep Learning (DL) Framework 2698.4 Applications 2718.4.1 Microwave Imaging of Free-Space and Buried Objects 2718.4.2 Biomedical Imaging 2728.4.3 Non-destructive Testing and Evaluation (NDT/NDE) 2748.4.4 Wireless Detection, Localization, and Tracking of Targets 2758.5 Conclusions 276Acknowledgments 276References 2779 RA-Based RIS-1 Design Using Support Vector Machines to Enhance mmWave 5G Coverage 283Álvaro F.Vaquero, Eduardo Martinez-de-Rioja, Jesús A. López-Fernández, and Manuel Arrebola9.1 Introduction 2839.1.1 RA-Based Reflective Intelligent Surface 2859.1.2 Considerations of RA-Based RIS Design 2879.2 RIS-1 Unit-Cell Characterization Using SVR 2899.2.1 Passive Unit Cell for RIS-1 Design 2899.2.2 SVR-Based Models of RA Unit Cells 2919.2.2.1 SVM Theoretical Background 2939.2.2.2 Model Selection, Expected Accuracy, and Training 2979.2.2.3 Efficient Grid Search 2999.3 RIS-1: Analysis and Optimization 3029.3.1 Radiated Field by a RIS 3049.3.1.1 Electric Field on the RIS Aperture 3049.3.1.2 Radiated Field of an RIS 3079.3.2 Intersection Approach Framework 3119.3.3 Generalized Intersection Approach 3159.4 SVR-Based Design of RIS-1 to Enhance 5G mmWave NF Coverage 3179.4.1 Definition of Scenario and Single-Layer Unit Cell 3179.4.2 Unit-Cell Modeling Based on SVR 3209.4.2.1 Discussion on the Number of Training Patterns, Time Cost and Achieved Precision 3219.4.2.2 Reflection Coefficients 3239.4.3 RIS-1 Designed Based on Intersection Approach Framework 3259.4.4 RIS-1 Design Process 3299.5 Conclusions and Road Map 332References 33410 AI at the Physical Layer for Wireless Network Security and Privacy 341Aly S. Abdalla, Bo Tang, and Vuk Marojevic10.1 Introduction 34110.2 Network Security and Privacy Threats and Vulnerabilities 34210.2.1 Security Threats 34210.2.2 Identifying and Assessing Network Security and Privacy Threats 34310.2.3 Exploiting Vulnerabilities: Techniques and Attack Vectors 34410.3 Fundamentals of AI for Network Security and Privacy 34610.3.1 Supervised Learning 34710.3.2 Unsupervised Learning 34910.3.3 Reinforcement Learning 35010.3.4 Generative Adversarial Networks 35110.3.5 Federated Learning 35210.3.6 Ensemble Learning 35310.4 AI-Driven Physical Layer Security Solutions 35510.4.1 Intelligent Beamforming 35610.4.2 AI-Based Radio Frequency Fingerprinting Techniques 35710.4.3 AI-Assisted Power Control 35810.5 Case Study: UAV-Assisted PLS for Terrestrial Wireless Communications Networks 35910.6 Practical Considerations and Challenges of Implementing AI-Based Security Solutions 36610.6.1 Scalability and Performance Optimization of AI Models 36610.6.2 Privacy Considerations of AI-Enhanced Wireless Network Security 36710.7 Conclusions and Outlook 369References 370Index 381