Artificial Intelligence for Future Networks
Description
An exploration of connected intelligent edge, artificial intelligence, and machine learning for B5G/6G architecture
Artificial Intelligence for Future Networks illuminates how artificial intelligence (AI) and machine learning (ML) influence the general architecture and improve the usability of future networks like B5G and 6G through increased system capacity, low latency, high reliability, greater spectrum efficiency, and support of massive internet of things (mIoT).
The book reviews network design and management, offering an in-depth treatment of AI oriented future networks infrastructure. Providing up-to-date materials for AI empowered resource management and extensive discussion on energy-efficient communications, this book incorporates a thorough analysis of the recent advancement and potential applications of ML and AI in future networks.
Each chapter is written by an expert at the forefront of AI and ML research, highlighting current design and engineering practices and emphasizing challenging issues related to future wireless applications.
Some of the topics include:
- Signal processing and detection, covering preprocess and level signals, transform signals and extract features, and training and deploying AI models and systems
- Channel estimation and prediction, covering channel characteristics, modeling, and classic learning-aided and AI-aided estimation techniques
- Resource allocation, covering resource allocation optimization and efficient power consumption for different computing paradigms such as Cloud, Edge, Fog, IoT, and MEC
- Antenna design using AI, covering basics of antennas, EM simulator/optimization algorithms, and surrogate modeling
Identifying technical roadblocks and sharing cutting-edge research on developing methodologies, Artificial Intelligence for Future Networks is an essential reference on the subject for professionals and researchers involved in the field of wireless communications and networks, along with graduate and PhD students in electrical and computer engineering programs of study.
More details
Other editions
Additional editions
Persons
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.
Content
About the Editors xv
List of Contributors xvii
Acknowledgments xxi
1 Intelligent Beam Prediction and Tracking 1
Christos Masouros, Jianjun Zhang, and Yongming Huang
1.1 Introduction 1
1.2 Challenge of Beam Prediction Modeling in Wireless Communications 5
1.3 Prior Identification - Perspective of Function Space 7
1.3.1 Perspective of Function Space 8
1.3.2 Useful Priors for Beam Process Modeling 9
1.3.2.1 High-speed Train Communication 9
1.3.2.2 Indoor Environment 9
1.3.2.3 City Street Environment 9
1.4 Methodology from Stochastic Process 12
1.5 Stochastic Continuity - Beam Index Difference 16
1.5.1 Beam Index Difference Technique 16
1.5.2 BPT Solution via Beam Index Difference 17
1.5.3 Theoretical Analysis for Beam Index Difference 21
1.6 Stochastic Smoothness - Hybrid Data-induced Kalman Filtering 25
1.6.1 Theoretical Foundation 26
1.6.2 Implicit Dynamics Learning via Multitask Learning 28
1.6.3 SDE Representation and Efficient Inference 31
1.7 Beam Width Optimization 33
1.7.1 Stochastic Continuity - Locality Principle of Beam Change and Data Transmission with Multiresolution Beam 33
1.7.2 Stochastic Smoothness - Low-frequency Sounding via BWO and Long-term Prediction 35
1.8 Numerical Results 36
1.8.1 Simulation Results for Stochastic Continuity 37
1.8.2 Simulation Results for Stochastic Smoothness 39
1.9 Conclusion 45
References 46
2 Signal Detection with Machine Learning 51
Jayakrishnan Vijayamohanan, Arjun Gupta, Manel Martínez-Ramón, and Christos Christodoulou
2.1 Introduction 51
2.2 Symbol Detection 52
2.2.1 The Viterbi Algorithm 52
2.2.2 Channel Equalization Through Machine Learning 54
2.2.3 Machine Learning Implementations of the Viterbi Algorithm 57
2.3 Modulation Detection 60
2.3.1 Signal Model 61
2.3.2 Feature Selection 62
2.3.3 Maximum Likelihood Estimation 64
2.3.4 Neural Modulation Detection 64
2.3.4.1 Convolutional Neural Network 65
2.3.4.2 CNN Modulation Detection 67
2.4 Source Detection 74
2.4.1 Array Signal Model 74
2.4.2 Conventional Source Detection 77
2.4.3 Neural Source Detection 79
2.4.3.1 CNN Detector 80
2.4.3.2 RadioNet 82
2.5 Conclusion 84
References 85
3 AI-Aided Channel Prediction 93
Oscar Stenhammar, Gábor Fodor, and Carlo Fischione
Acronyms 93
3.1 Introduction 94
3.1.1 Channel Aging 94
3.1.2 Channel Estimation 96
3.1.3 Channel Prediction 96
3.2 Preliminaries 98
3.2.1 Multilayer Perceptron 98
3.2.2 Convolutional Neural Network 100
3.2.3 Recurrent Neural Network 101
3.2.3.1 Long Short-Term Memory 101
3.2.3.2 Gated Recurrent Units 103
3.2.4 Transformer 103
3.3 Previous Work 105
3.3.1 Previous Work in Channel Estimation 105
3.3.2 Conventional Channel Prediction 107
3.3.3 Previous Work in AI-Aided Channel Prediction 109
3.4 Experimental Evaluations 113
3.4.1 Simulation Setup 113
3.4.2 Neural Network Setup 115
3.4.3 Experimental Results 118
3.5 Discussion 121
3.6 Summary 123
References 124
4 Semantic Communications 131
Qiyang Zhao, Hang Zou, Mehdi Bennis, and Merouane Debbah
4.1 Introduction 131
4.2 Semantic Information and Semantic-Native Communication 134
4.2.1 Semantic Information Theory 134
4.2.2 Semantic-Native Communication 137
4.3 Interplay of AI and Semantic Communication 140
4.3.1 AI for Semantic Communication 140
4.3.2 Semantic-Native Collective Intelligence 143
4.4 Conclusion 145
References 146
5 Federated Learning for Wireless Communications 151
Ahmet M. Elbir and Wei Shi
5.1 Introduction 151
5.2 Channel Models 155
5.2.1 mmWave Channel Model 155
5.2.2 THz Channel Model 157
5.2.2.1 Near-Field Array Model 158
5.2.2.2 Near-Field Beam Squint 160
5.3 Federated Learning for Channel Estimation 162
5.3.1 Training Data Collection 162
5.3.2 FL-Based Model Training 163
5.3.3 FL for mmWave Channel Estimation in Massive MIMO 165
5.3.4 FL for mmWave Channel Estimation in RIS-Assisted Massive Mimo 169
5.3.5 FL for THz Channel Estimation 172
5.4 FL For Hybrid Beamforming 176
5.5 Conclusions 178
Acknowledgment 179
References 179
6 Federated Learning in Mesh Networks 185
Xu Wang, Yuanzhu Chen, and Octavia A. Dobre
6.1 Introduction 185
6.1.1 Federated Learning 185
6.1.2 Mesh Networks 186
6.1.3 The Convergence: Federated Learning on Mesh Networks 187
6.2 Decentralized Federated Learning 188
6.2.1 Traditional Federated Learning versus Decentralized Federated Learning 189
6.2.2 Core Principles of Decentralized Federated Learning 191
6.2.3 Advantages of Decentralization in Federated Learning 191
6.2.4 Architecture Variants for Decentralized Federated Learning 192
6.2.5 Challenges of Decentralization in Federated Learning 192
6.3 Mesh Networks 192
6.3.1 Why Mesh Networks 193
6.3.2 Fundamental Concepts and Terminologies 193
6.3.3 Topological Structures 193
6.3.4 Advantages of Mesh Networks 194
6.3.5 Challenges and Limitations 195
6.3.6 Integration with Federated Learning 195
6.4 The Intersection: Decentralized Federated Learning over Mesh Networks 196
6.4.1 Natural Synergy Between Federated Learning and Mesh Networks 196
6.4.2 Potential Benefits of the Convergence 196
6.4.3 Enabling Technologies 198
6.4.4 Challenges at the Intersection 198
6.4.4.1 Communication Overhead 198
6.4.4.2 Data Heterogeneity and Non-IID Data 199
6.4.4.3 Model Aggregation in Decentralized Networks 199
6.4.4.4 Network Latency and Asynchrony 199
6.4.4.5 Security and Privacy Concerns 199
6.4.4.6 Scalability Concerns 200
6.4.4.7 Fault Tolerance and Robustness 200
6.4.4.8 Resource Constraints 200
6.5 Solutions 200
6.5.1 Communication Overhead 200
6.5.2 Data Heterogeneity and Non-IID Data 201
6.5.3 Model Aggregation in Decentralized Networks 201
6.5.4 Latency and Asynchrony 202
6.5.5 Security and Privacy Concerns 202
6.5.6 Scalability Concerns 202
6.5.7 Fault Tolerance and Robustness 203
6.5.8 Resource Constraints 203
6.6 State-of-the-Art and Noteworthy Implementations 204
6.6.1 Decentralized Federated Learning Techniques 204
6.6.1.1 Network Topology 204
6.6.1.2 Communication Protocols 204
6.6.1.3 Privacy Enhancements 205
6.6.2 Advances in Mesh Networking Technologies 205
6.6.2.1 Low-Latency Protocols 205
6.6.2.2 Scalable Architectures 206
6.6.2.3 Security Enhancements 206
6.6.3 Decentralized Federated Learning on Mesh Networks: Integrated Approaches 206
6.6.4 Toolkits and Platforms 207
6.6.5 Benchmarks and Evaluation 208
6.7 Future Directions and Open Research Challenges 209
6.7.1 Advanced Algorithms 209
6.7.2 Enhanced Security Mechanisms 209
6.7.3 Network Optimization 210
6.7.4 Interoperability and Standardization 210
6.7.5 Energy Efficiency and Sustainability 211
6.7.6 User-Centric Approaches 211
6.7.7 Real-time Decentralized Federated Learning 212
6.7.8 Codesigning Hardware and Software 212
6.7.9 Ethical and Regulatory Considerations 213
6.7.10 Interdisciplinary Research 213
6.8 Concluding Remarks 213
References 214
7 Antenna Design Using Artificial Intelligence 227
Sotirios K. Goudos, Mohammad A. Matin, and George K. Karagiannidis
7.1 Introduction 227
7.2 Evolutionary Algorithms 229
7.2.1 Mainstream Algorithms 229
7.2.1.1 Genetic Algorithms 229
7.2.1.2 Particle Swarm Optimization 230
7.2.1.3 Differential Evolution 231
7.2.1.4 Ant Colony Optimization 232
7.2.2 Emerging Algorithms 235
7.2.2.1 Biogeography-Based Optimization 235
7.2.2.2 Grey Wolf Optimizer 235
7.2.2.3 Wind-Driven Optimization 235
7.2.2.4 Salp Swarm Algorithm 235
7.2.2.5 Artificial Bee Colony (ABC) 236
7.2.2.6 Harmony Search (HS) 236
7.2.2.7 Shuffled Frog-Leaping Algorithm 237
7.2.3 Antenna Optimization Using Evolutionary Algorithms 237
7.2.3.1 Problem Formulation 237
7.2.3.2 Numerical Results 239
7.3 Machine Learning 244
7.3.1 Artificial Neural Networks (ANNs) 244
7.3.2 Support Vector Machines 244
7.3.3 Gaussian Process (GP) 245
7.3.4 Deep Learning (DL) 245
7.3.5 ANFIS 245
7.3.6 Surrogate Modeling 246
7.3.6.1 Surrogate Modeling Example 248
7.4 Knowledge Representation 252
7.5 Conclusion 253
References 253
8 AI-Driven Approaches for Solving Electromagnetic Inverse Problems 257
Marco Salucci, Maokun Li, and Andrea Massa
8.1 Introduction 257
8.2 Mathematical Formulation 258
8.3 AI-Based EM-IP Solution Strategies 262
8.3.1 3-Step Learning-by-Examples (LBE) Framework 263
8.3.2 System-by-Design (SbD) Framework 267
8.3.3 Deep Learning (DL) Framework 269
8.4 Applications 271
8.4.1 Microwave Imaging of Free-Space and Buried Objects 271
8.4.2 Biomedical Imaging 272
8.4.3 Non-destructive Testing and Evaluation (NDT/NDE) 274
8.4.4 Wireless Detection, Localization, and Tracking of Targets 275
8.5 Conclusions 276
Acknowledgments 276
References 277
9 RA-Based RIS-1 Design Using Support Vector Machines to Enhance mmWave 5G Coverage 283
Álvaro F.Vaquero, Eduardo Martinez-de-Rioja, Jesús A. López-Fernández, and Manuel Arrebola
9.1 Introduction 283
9.1.1 RA-Based Reflective Intelligent Surface 285
9.1.2 Considerations of RA-Based RIS Design 287
9.2 RIS-1 Unit-Cell Characterization Using SVR 289
9.2.1 Passive Unit Cell for RIS-1 Design 289
9.2.2 SVR-Based Models of RA Unit Cells 291
9.2.2.1 SVM Theoretical Background 293
9.2.2.2 Model Selection, Expected Accuracy, and Training 297
9.2.2.3 Efficient Grid Search 299
9.3 RIS-1: Analysis and Optimization 302
9.3.1 Radiated Field by a RIS 304
9.3.1.1 Electric Field on the RIS Aperture 304
9.3.1.2 Radiated Field of an RIS 307
9.3.2 Intersection Approach Framework 311
9.3.3 Generalized Intersection Approach 315
9.4 SVR-Based Design of RIS-1 to Enhance 5G mmWave NF Coverage 317
9.4.1 Definition of Scenario and Single-Layer Unit Cell 317
9.4.2 Unit-Cell Modeling Based on SVR 320
9.4.2.1 Discussion on the Number of Training Patterns, Time Cost and Achieved Precision 321
9.4.2.2 Reflection Coefficients 323
9.4.3 RIS-1 Designed Based on Intersection Approach Framework 325
9.4.4 RIS-1 Design Process 329
9.5 Conclusions and Road Map 332
References 334
10 AI at the Physical Layer for Wireless Network Security and Privacy 341
Aly S. Abdalla, Bo Tang, and Vuk Marojevic
10.1 Introduction 341
10.2 Network Security and Privacy Threats and Vulnerabilities 342
10.2.1 Security Threats 342
10.2.2 Identifying and Assessing Network Security and Privacy Threats 343
10.2.3 Exploiting Vulnerabilities: Techniques and Attack Vectors 344
10.3 Fundamentals of AI for Network Security and Privacy 346
10.3.1 Supervised Learning 347
10.3.2 Unsupervised Learning 349
10.3.3 Reinforcement Learning 350
10.3.4 Generative Adversarial Networks 351
10.3.5 Federated Learning 352
10.3.6 Ensemble Learning 353
10.4 AI-Driven Physical Layer Security Solutions 355
10.4.1 Intelligent Beamforming 356
10.4.2 AI-Based Radio Frequency Fingerprinting Techniques 357
10.4.3 AI-Assisted Power Control 358
10.5 Case Study: UAV-Assisted PLS for Terrestrial Wireless Communications Networks 359
10.6 Practical Considerations and Challenges of Implementing AI-Based Security Solutions 366
10.6.1 Scalability and Performance Optimization of AI Models 366
10.6.2 Privacy Considerations of AI-Enhanced Wireless Network Security 367
10.7 Conclusions and Outlook 369
References 370
Index 381
1
Intelligent Beam Prediction and Tracking
Christos Masouros1, Jianjun Zhang2, and Yongming Huang3
1Department of Electronic & Electrical Engineering, University College London, London, UK
2College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing, China
3National Mobile Communications Research Laboratory, Southeast University, Nanjing, China
CHAPTER MENU
- Introduction
- Challenge of Beam Prediction Modeling in Wireless Communications
- Prior Identification - Perspective of Function Space
- Methodology from Stochastic Process
- Stochastic Continuity - Beam Index Difference
- Stochastic Smoothness - Hybrid Data-induced Kalman Filtering
- Beam Width Optimization
- Numerical Results
- Conclusion
1.1 Introduction
Because of abundant spectrum resources at high-frequency band, that enable to achieve ultrahigh-speed data transmission (DT), high-frequency communications, e.g., millimeter wave or even Terahertz communications, have attracted extensive interest from academia, industry, and government [1]. For high-frequency communications, the transmitter and/or receiver are often equipped with large-scale antenna arrays, i.e., massive multiple-input multiple-output (MIMO), to achieve high array gains to overcome signal attenuation of high-frequency band. However, the use of pencil-like highly directional beams makes channel state information (CSI) acquisition via beam alignment (BA) very challenging. First, acquiring CSI in a mobile network is particularly challenging since the wireless channel often varies rapidly. Second, in contrast to the fully digital transceiver, where the pilot transmission scheme can be utilized to acquire CSI [2, 3], channel estimation is more complicated in the hybrid antenna array architecture, which has been used widely in practice, because we cannot extract the actual received signals on all antennas simultaneously. Last but not the least, the large dimension of massive MIMO inevitably results in large and even unaffordable pilot overhead, even if the pilot transmission-based method can be used.
To tackle this challenging issue, the two-step precoding and combining based scheme is widely used in practical systems, e.g., standardized in IEEE 802.11ad/802.15.3c [4, 5]. Let , , and represent the precoding matrix, channel matrix, and combing matrix, respectively. The precoding (combining) matrix is assumed to be decomposed as (), where and ( and ) denote the analog and digital parts, respectively. To reduce the dimension of the CSI estimate problem, beam training is first performed between the transmitter and receiver to obtain the analog precoder and combiner . Then, the effective or equivalent channel can be estimated, based on which the digital parts, i.e., the precoder and combiner , can be designed in the analog domain via a variety of methods, e.g., the heuristic methods or optimization-based algorithms [6-8]. Note that since the size of the effective channel matrix is much smaller than that of the original channel matrix , the pilot overhead in the second step is relatively low. It is observed that the remaining difficulty lies in how to design an efficient beam training scheme to find the optimal and .1
Initially, beam training (also referred to as beam sounding) is implemented via the exhaustive and hierarchical search [5, 9, 10]. Compared to the exhaustive search, whose sounding overhead is with denoting the size of the training codebook, the sounding overhead of the hierarchical search is for the typical binary tree search-based implementation, which is smaller than that of the exhaustive search scheme. For this reason, along with the advantage of easy implementation, the hierarchical search-based scheme has been adopted in several IEEE standards, such as IEEE 802.15.3c and IEEE 802.11aj. Note the performance of the hierarchical search-based algorithms heavily depends on the codebook used. In fact, besides the demand for multi resolution, namely, various widths of main lobes, other properties, such as flat main lobe and side lobe, narrow transition band, and high-power efficiency of power amplifier, are also very important and should be well addressed [9, 11]. In general, the research on hierarchical search often boils down to sounding codebook design [5, 9-14].
The advantage of the exhaustive or hierarchical search-based methods is that they can be applied to an arbitrary scenario because they are nonadaptive methods and thus independent of external environments. However, the beam sounding overhead is almost always very large, especially for a large-scale antenna array and/or a rapidly changing environment. In fact, on the one hand, as the scale of the antenna array increases, the beam width decreases accordingly, which thus increases the sounding overhead. On the contrary other hand, the coherence time or period becomes shorter in a rapidly fluctuating environment. Hence, much of the precious time resource is spent on beam sounding, and the proportion of time resources used for DT is very small. This phenomenon is particularly pronounced for the highly varying communication scenarios, e.g., unmanned aerial vehicle (UAV) communication.
To avoid frequent searches, beam tracking is invoked to reduce the sounding overhead. The complete process of BA operation in a relatively long time consists of two phases. First, initial BA is performed in the first stage to find the optimal beam or beam pair via the exhaustive or hierarchical search, which involves a large beam sounding overhead, as mentioned before. Then, the beam tracking technique is invoked in the second phase to enable efficient search. Compared to the initial BA, the number of beams used for tracking is relatively small, e.g., maybe only one beam is used for sounding. Note that if the tracking fails, which is inevitable, the initial BA operation is invoked again to reinitialize the beam tracking.
The key to beam tracking is beam prediction, i.e., to predict a beam subspace that contains the real beam. In practice, two types of metrics are closely related to beam prediction. The first one is the success rate and prediction efficiency, i.e., the beam subspace predicted should contain the real optimal beam, and meanwhile, the beam subspace should be as small as possible. The second one is the complexities of beam prediction, including both sample complexity and inference complexity. To balance these indicators, various methods have been proposed, the core of which is to exploit temporal and spatial correlations of wireless channels. The most important step toward beam prediction is to construct an appropriate prediction model. Overall, there are mainly two ways to construct a prediction model, i.e., the traditional manual fashion and the recent automatic fashion. The classical and representative manual method to construct a prediction model is the Kalman filtering- or Bayesian filtering-based Beam prediction and tracking (BPT) algorithms [15-22]. Machine learning (ML) methods are used to automatically construct prediction models, typically, in the data-driven manner [23-29].
The Kalman filtering and Bayesian filtering methods address the issue of prediction model construction by building a dynamical model that characterizes the underlying physical system. Specifically, two stochastic differential equations (SDEs), referred to as state-space and measurement equations in literature, are first established. As long as the two SDEs are available, the well-known Kalman filter or Bayesian filter can be invoked to perform real-time inference or prediction. For example, both the extended Kalman filter- and Bayesian filter-based beam tracking algorithms are proposed in Liu et al. [15] and Yuan et al. [18] for the dual-functional radar and communication systems. For the distributed millimeter-wave massive MIMO problem, a monopulse beam tracking method based on the unscented Kalman filter is designed in [21], which shows to achieve good robustness as well as generalization ability.
An important and appealing advantage of the Kalman filtering based methods is that they have low computational complexity. In particular, the scaling of computational complexity for the Kalman filter is linear (where is the number of samples), as opposed to the cubic scaling for Gaussian process (GP) regression-based BPT algorithms. Note that this advantage is attributed to the fact that the underlying state-space system model is exploited. However, since the prediction model is obtained via manual derivation manner, it may fail in complicated scenarios or environments. To tackle this issue, recently a novel hybrid model and data-driven-based approach, referred to as hybrid data-induced Kalman filtering (HDIKF), has been proposed by Zhang et al. [30, 31].
In contrast to the Kalman filtering-based designs, ML typically addresses the issue of prediction modeling by employing the data-driven mode. In fact, it is well known that a powerful ability of ML is that it can automatically extract meaningful patterns and further derive directly an appropriate model from the observed data. According to the underlying ML theory and methods, the ML-based beam prediction methods fall into two categories, i.e., the (deep) reinforcement learning...
