Broadband Communications, Computing, and Control for Ubiquitous Intelligence

Preface
Contents
Contributors
1 Tribute to Professor Jon W. Mark
Personal Stories
Greeting Messages from Alumni
Part I Broadband Communications for Ubiquitous Connectivity
2 Network Slicing for 5G Networks and Beyond
2.1 Introduction to 5G Communication Networks
2.2 Network Slicing
2.2.1 Network Slicing in 5G Wireless Networks
2.2.1.1 Dynamic Radio Resource Slicing Framework
2.2.2 Network Slicing in 5G Core Networks
2.2.2.1 Joint Computing and Transmission Resource Slicing
2.2.3 AI-Assisted Network Slicing in Beyond 5G Networks
2.2.3.1 Beyond 5G Networks
2.2.3.2 AI-Assisted Network Slicing
2.3 Case Study
2.4 Conclusion
References
3 Responsive Regulation of Dynamic UAV Communication Networks Based on Deep Reinforcement Learning
3.1 Introduction
3.2 Related Works
3.3 System Model and Problem Formulation
3.3.1 Network Environment
3.3.2 Spectrum Access
3.3.3 Energy-Related Considerations
3.3.4 Problem Formulation
3.4 Preliminaries
3.5 Learning Algorithm Design for Proactive Self-Regulation Strategy
3.5.1 State Space
3.5.1.1 Case of UAV Quit
3.5.1.2 Case of UAV Join-In
3.5.2 Action Definition
3.5.3 Reward Function Design
3.5.4 State Transition Definition
3.5.4.1 Case of UAV Quit
3.5.4.2 Case of UAV Join-In
3.5.5 Training Tune-Ups
3.5.5.1 Tune-Ups for Neural Network Training
3.5.5.2 Tune-Ups for RL Training
3.5.6 Parallel Computing
3.6 Proactive Self-Regulation with Dynamic User Distribution
3.7 Numerical Results
3.7.1 Simulation Setup
3.7.2 Simulation Results
3.7.2.1 Case Without UAV or User Dynamics
3.7.2.2 Case of UAV Quit
3.7.2.3 Case of UAV Join-In
3.7.2.4 Case of UAV and User Dynamics
3.8 Conclusions
References
4 Utility-Based Dynamic Resource Allocation in IEEE 802.11ax Networks: A Genetic Algorithm Approach
4.1 Introduction
4.2 Related Works
4.3 Background on OFDMA and RU Allocation in IEEE 802.11ax
4.4 System Model
4.5 Utility-Based Dynamic Resource Allocation Scheme
4.5.1 Optimal Resource Allocation Problem Formulation
4.5.2 Genetic Algorithm
4.6 Simulation Results
4.6.1 UDRA vs. Exhaustive Search
4.6.2 Network-Wise Throughputs and Fairness Indexes
4.7 Conclusion
References
5 Intelligentized Radio Access Network for Joint Optimization of User Association and Power Allocation
5.1 Introduction
5.2 Related Work
5.3 Main Contribution
5.4 System Model
5.5 Problem Formulation
5.6 DQL Framework
5.6.1 DQN
5.6.2 Design the DQN
5.6.2.1 Actions
5.6.2.2 Reward
5.7 Results and Discussions
5.7.1 Training and Testing Results
5.7.2 UE Performance
5.7.3 Robustness
5.7.4 Scalability
5.7.5 Closer Look at DQN
5.8 Summary
References
6 Routing Algorithms for Heterogeneous Vehicular Networks
6.1 Introduction
6.2 Background
6.2.1 Unicast Routing Algorithms
6.2.2 Broadcast Routing Algorithms
6.2.3 Geocast Routing Algorithms
6.2.4 Related Work in Routing Algorithms
6.3 Machine Learning-Based Routing Algorithm for IoV with Mobility Prediction
6.3.1 Network Model
6.3.2 Statistical Mobility Model
6.3.2.1 Inter-Arrival Time Distribution
6.3.2.2 Inter-Vehicle Spacing Distribution
6.3.3 Channel Model
6.3.4 ANN Model
6.4 Performance Evaluation
6.5 Conclusion
References
7 Teaching from Home: Computer and Communication Network Perspectives
7.1 Introduction
7.2 Related Work
7.3 Network Technologies Involved
7.3.1 Host Computers
7.3.1.1 Desktop, Laptop, or Tablet?
7.3.1.2 Windows, Mac OS, or Linux?
7.3.1.3 Other Necessary Peripherals
7.3.2 Home Networks
7.3.2.1 Ethernet Structured Wiring
7.3.2.2 No-New-Wires Home Backbone
7.3.2.3 Wireless Home Network
7.3.3 Internet Access
7.3.3.1 Fiber, Cellular, or Satellite?
7.3.3.2 Telephone Service Providers
7.3.3.3 Television Service Providers
7.4 Improvement for Online Teaching
7.4.1 WiFi Interference Avoidance
7.4.1.1 A Better (Al)located WiFi AP
7.4.1.2 Wired Interconnected WiFi APs
7.4.1.3 Wireless Interconnected WiFi APs
7.4.2 WAN Reliability Augmentation
7.4.2.1 DSL vs. Cable Modem
7.4.2.2 Primary vs. Backup
7.4.2.3 Load Balancing
7.4.3 Recommendations on Teaching from Home
7.5 Further Discussion
7.6 Conclusions
References
Part II Caching, Computing, and Control for Ubiquitous Intelligence
8 State Transition Field: A New Framework for Mobile Dynamic Caching
8.1 Introduction
8.2 State Transition Field
8.2.1 Content Request and Replacement
8.2.2 Cache State
8.2.3 State and Content Caching Probabilities
8.2.4 General Cache State Transition Model
8.2.5 State Transition Field
8.2.6 Discussions on the Steady State and the Convergence
8.3 State Transition Field with Time-Varying Content Popularity
8.3.1 General Replacement Model
8.3.2 Instantaneous STF: The General Case
8.3.3 Impact of STF on Instantaneous Cache Hit Probability
8.4 Dynamic Probabilistic Caching with Time-Varying Content Popularity
8.4.1 The Content Replacement Markov Chain
8.4.2 Generating the State Transition Matrix
8.4.3 Discussion on Scalability
8.5 Numerical Results
8.5.1 State Transition Field with Time-Invariant Content Popularity
8.5.2 State Transition Field with Time-Varying Content Popularity
8.5.3 Dynamic Probabilistic Caching with Time-Varying Content Popularity
8.6 Summary
References
9 Deep Reinforcement Learning for Mobile EdgeComputing Systems
9.1 Introduction
9.2 Overview of Deep Reinforcement Learning
9.2.1 DRL Problem Formulation
9.2.2 Determine the Optimal Policy with Deep Learning
9.2.3 Existing DRL Algorithms
9.3 Case Study: Deep Q-Learning for Task Offloading in MEC
9.3.1 System Model
9.3.1.1 Task Model
9.3.1.2 Task Offloading Decision
9.3.1.3 Local Processing Model
9.3.1.4 Edge Node Offloading Model
9.3.2 Task Offloading Problem
9.3.2.1 State
9.3.2.2 Action
9.3.2.3 Cost
9.3.2.4 Problem Formulation
9.3.3 Deep Q-Learning-Based Algorithm
9.3.3.1 Neural Network
9.3.3.2 Algorithm Design
9.3.4 Performance Evaluation
9.3.4.1 Algorithm Convergence
9.3.4.2 Method Comparison
9.4 Challenges and Future Directions
9.5 Conclusion
References
10 Mobile Computation Offloading with Hard TaskCompletion Times
10.1 Introduction
10.2 Continuous Offloading
10.2.1 System Description and Problem Formulation
10.2.1.1 Local Execution
10.2.1.2 Remote Execution
10.2.2 Markovian Channel and the Time-Dilated Absorbing Markov Model
10.2.3 Offline Bound
10.2.4 OnOpt (Online Optimal) Algorithm
10.3 Multi-part Offloading
10.3.1 Problem Formulation
10.3.2 Offline Bound
10.3.3 The Time-Dilated Absorbing Markov Model
10.3.4 Optimal Algorithm for K-Part Offloading
10.4 Numerical Results
10.4.1 Simulation Set 1
10.4.2 Simulation Set 2
10.5 Summary
References
11 Online Incentive Mechanism Design in Edge Computing
11.1 Introduction
11.2 Mechanism Design and Auction
11.3 Primal–Dual-Based Online Incentive Mechanism
11.3.1 Primal–Dual-Based Method for Linear Systems
11.3.2 Primal–Dual-Based Method for Nonlinear Systems
11.4 Application of Primal–Dual Online Incentive Mechanism Design in Edge Computing
11.4.1 System Model Descriptions
11.4.1.1 System Model
11.4.1.2 Problem Formulation
11.4.2 The Design of OMAP
11.4.2.1 Problem Reformulation
11.4.2.2 OMAP
11.4.3 Performance Analyses
11.4.4 Numerical Simulations
11.5 Summary
References
12 Collaborative Deep Neural Network Inference via Mobile Edge Computing
12.1 Introduction
12.2 Background
12.2.1 DNN Inference
12.2.2 Mobile Edge Computing
12.2.3 Machine Learning
12.3 Collaborative DNN Inference via Device-Edge Orchestration
12.3.1 Collaborative DNN Inference Framework
12.3.2 Service Delay and Accuracy Analysis of Collaborative DNN Inference
12.3.2.1 Inference Delay Analysis
12.3.2.2 Inference Accuracy Analysis
12.3.3 Joint Sampling Rate Selection and Resource Allocation Problem
12.3.3.1 Constrained Markov Decision Process
12.3.4 Deep RL-Based Solution
12.3.4.1 Markov Decision Process Transformation (Step 1)
12.3.4.2 Optimization Subroutine for Resource Allocation (Step 3)
12.3.4.3 Deep RL-Based Algorithm (Step 2)
12.4 Performance Evaluation
12.4.1 Experiment Setup
12.4.2 Convergence Performance
12.4.3 Impact of Task Arrival Rate
12.4.4 Impact of Optimization Subroutine
12.5 Conclusion
References
13 Automated Data-Driven System for Compliance Monitoring
13.1 Introduction
13.1.1 Radio Spectrum Management
13.1.2 Spectrum Monitoring for Compliance
13.1.3 Chapter Contributions and Organization
13.2 Automated Data-driven System
13.3 Data Sources
13.3.1 Spectrum Measurements
13.3.2 Spectrum Management Records
13.4 Signal Identification
13.4.1 Mode Analysis
13.4.2 Mode-Sensor Matching
13.4.3 License-Measurements Association
13.5 Violation Identification
13.5.1 Detecting Violations
13.5.2 Characterizing Violations
13.5.2.1 Confidence Indicators
13.5.2.2 Behavioral Indicators
13.5.2.3 Extent Indicators
13.5.2.4 Impact Indicators
13.5.3 Prioritizing Violations
13.6 Summary
References
14 AI Driven User Authentication
14.1 Introduction
14.2 Facial Recognition
14.2.1 Overview
14.2.2 Facial Recognition Using EigenFace Algorithm
14.2.3 Facial Recognition Using CNN
14.3 Implementation
14.3.1 Mobile Authenticator
14.3.2 Supporting Cloud Backend
14.4 Conclusion
References
15 Control and Communication Coordination for Industrial Digital Twins of Sintering Process
15.1 Introduction
15.2 Control–Communication Coordination Architecture for Industrial Digital Twins
15.3 Sintering Production Line
15.4 Deterministic Communication Based on Time-Sensitive Networking
15.5 Intelligent Modeling for Sintering Process
15.6 Digital Twins Coordination of the Sintering Process
15.7 Summary
References
Index