Industrial Edge Computing : Architecture, Optimization and Applications

✍ Scribed by Xiaobo Zhou; Shuxin Ge; Jiancheng Chi; Tie Qiu

• Publisher: Springer Nature Singapore
• Year: 2024
• Tongue: English
• Leaves: 216
• Category: Library

✦ Synopsis

This book serves as a pivotal guide for professionals and researchers within the industrial computing domain, offering an extensive examination of edge computing in industrial environments. Tailored for individuals possessing a foundational understanding of industrial computing systems, it aims to augment their knowledge concerning the role and capabilities of edge computing in this dynamically evolving sector.

In an era where real-time, reliable, and scalable computing solutions are of paramount importance, traditional cloud computing models grapple with challenges such as latency, bandwidth limitations, data sovereignty, and privacy concerns. This book positions edge computing as a crucial evolution in industrial data processing and analytics, specifically addressing these challenges. It introduces a distinctive three-layer industrial edge computing architecture that integrates device, edge, and application layers, explicitly designed to accommodate the intricacies of the Industrial Internet of Things (IIoT).

Beyond elucidating the theoretical foundations of edge computing, the book delves into its practical applications, with a particular emphasize on edge-assisted model inference as a key scenario. It offers insightful case studies and discussions on the integration of edge computing with artificial intelligence (AI), illustrating how this collaboration is revolutionizing industrial systems. A comprehensive understanding of the material is facilitated by a background in computer science, industrial engineering, IoT, and cloud computing.

✦ Table of Contents

Preface
Contents
Acronyms
1 Introduction to Industrial Edge Computing
1.1 Concepts
1.1.1 What Is IIoT?
1.1.2 What Is Cloud Computing?
1.1.3 What Is Edge Computing?
1.2 Reference Architecture
1.2.1 Existing Reference Architectures
1.2.2 Proposed Reference Architectures
1.2.2.1 Device Layer
1.2.2.2 Edge Layer
1.2.2.3 Application Layer
1.3 Benefits and Challenges
1.3.1 Benefits of Industrial Edge Computing
1.3.2 Challenges of Industrial Edge Computing
1.3.2.1 5G-Based Edge Communication
1.3.2.2 Data Offloading and Load Balancing
1.3.2.3 Edge Artificial Intelligence
1.3.2.4 Data Sharing Security
1.4 Organization of This Book
References
2 Preliminaries
2.1 Performance Metrics of Industrial Edge Computing
2.1.1 Latency Minimization Scheme
2.1.2 Energy Consumption Trimming
2.1.3 Accelerating Inference with Improved Model Accuracy
2.2 Related Work
2.2.1 Offloading for Mass End Devices
2.2.1.1 From Decentralized to Centralized in Industrial Edge Computing
2.2.1.2 Hybrid Offloading
2.2.1.3 Offloading with Direct Acyclic Graph (DAG)-Based Partition
2.2.1.4 Offloading with Cooperation
2.2.2 Multisource Data Caching and Migration
2.2.2.1 Data Caching
2.2.2.2 Service Migration
2.2.3 Intelligent Application in Industrial Edge Computing
2.2.3.1 Inference Acceleration
2.2.3.2 Cooperative Inference for Improved Accuracy
References
3 Computation Offloading in Industrial Edge Computing
3.1 Introduction
3.2 Adaptive Offloading with Two-Stage Hybrid Matching
3.2.1 Statement of Problem
3.2.1.1 Overview
3.2.1.2 Communication Model
3.2.1.3 Computation Model
3.2.1.4 Problem Formulation
3.2.2 Scheme Overview
3.2.2.1 Global Buffer
3.2.2.2 Online Matching Stage
3.2.2.3 Offline Matching Stage
3.2.3 Global Buffer
3.2.4 Online Matching Stage
3.2.5 Offline Matching Stage
3.2.6 Performance Evaluation
3.2.6.1 Experiment Setup
3.2.6.2 Numerical Results
3.3 Dependent Offloading with DAG-Based Cooperation Gain
3.3.1 Statement of Problem
3.3.2 Cooperation Gain Estimation Based on DAG
3.3.3 Branch Soft Actor–Critic Offloading Algorithm
3.3.3.1 Problem Transformation
3.3.3.2 Soft Policy Function
3.3.3.3 Branch Soft Actor–Critic
3.3.4 Performance Evaluation
3.3.4.1 Impact of Bandwidth
3.3.4.2 Impact of Number of EDs
3.3.4.3 Impact of Cores
3.3.4.4 Impact of Weight Parameters
References
4 Data Caching in Industrial Edge Computing
4.1 Introduction
4.2 Freshness-Aware Caching with Distributed MAMAB
4.2.1 Statement of Problem
4.2.2 HD Map Caching Model
4.2.2.1 Overview
4.2.2.2 Vehicle Request Model
4.2.2.3 Specific Cost
4.2.3 Distributed Caching and Requesting Algorithm
4.2.3.1 Freshness-Aware Request
4.2.3.2 MAMAB-Based Caching
4.2.4 Performance Evaluation
4.3 Multicategory Video Caching
4.3.1 Statement of Problem
4.3.2 FoV-Based QoE of Users
4.3.3 Multi-agent Soft Actor–Critic Caching
4.3.4 Performance Evaluation
4.3.4.1 QoE Performance
4.3.4.2 Impact of the Proportion of Requests
4.3.4.3 Impact of Cache Size
References
5 Service Migration in Industrial Edge Computing
5.1 Introduction
5.2 Energy-Efficient Migration Based on 3-Layer VM Architecture
5.2.1 Statement of Problem
5.2.2 Energy-Efficient Service Migration Model Under 3-Layer VM Architecture
5.2.2.1 Service Latency
5.2.2.2 Energy Consumption
5.2.2.3 Problem Formulation
5.2.3 Lyapunov Optimization
5.2.4 Probabilistic Particle Swarm Optimization Algorithm
5.2.5 Performance Evaluation
5.2.5.1 Impact of User Number
5.2.5.2 Impact of BS Number
5.2.5.3 Impact of Duration τ
5.2.5.4 Impact of Mobility
5.2.5.5 Impact of V and V
5.2.5.6 Average Service Latency of Services
5.3 Location Privacy-Aware Service Migration
5.3.1 Statement of Problem
5.3.2 Adversary's Location Inference Attack
5.3.3 Location Privacy-Aware Multiuser Service Migration Algorithm
5.3.3.1 Problem Transformation
5.3.3.2 MASAC Service Migration Algorithm
5.3.4 Performance Evaluation
5.3.4.1 Impact of Wireless Bandwidth
5.3.4.2 Impact of Request Size
5.3.4.3 Impact of User Number
References
6 Application-Oriented Industrial Edge Computing
6.1 Image-Oriented Object Detection
6.1.1 Statement of Problem
6.1.2 Entry Point Selection
6.1.3 Computation Cost Estimation
6.1.4 Adaptive Offloading
6.1.5 Performance Evaluation
6.1.5.1 Object Detection Accuracy
6.1.5.2 Object Detection Latency
6.1.5.3 The Performance of Adaptive Sub-task Generation and Offloading
6.1.5.4 The Impact of Uniformly Sampled Zero-Padding Scheme
6.2 Point Cloud Oriented Object Detection
6.2.1 Statement of Problem
6.2.2 Point Cloud Partition
6.2.3 Data Alignment
6.2.4 Multilevel Data Fusion
6.2.5 K-Soft Actor–Critic Algorithm
6.2.6 Performance Evaluation
6.2.6.1 Evaluations on KITTI Dataset
6.2.6.2 Evaluation on Dataset Collected from Two Real Vehicles
6.3 Video Inference with Knowledge Distillation
6.3.1 Statement of Problem
6.3.2 Inference Accuracy Estimation
6.3.3 Cross Entropy Method (CEM)
6.3.4 Performance Evaluation
6.3.4.1 Overall Accuracy Improvement
6.3.4.2 Impact of BS Number
References
7 Future Research Directions
7.1 Theory Exploration for Future Directions
7.1.1 Digital Twin for Industrial Edge Computing System
7.1.2 Cloud–Edge Collaborative Data Analysis
7.1.3 Real-Time Communication with Less Data
7.2 Application Scenarios
7.2.1 Prognostics and Health Management
7.2.2 Smart Grid
7.2.3 Manufacturing
7.2.4 Intelligent Connected Vehicles
7.2.5 Smart Logistic
References