Distributed Edge Intelligence Enabled Wireless Communication Systems Serving Industrial Applications

Typ: Fortschritt-Berichte VDI
Erscheinungsdatum: 03.11.2022
Reihe: 08
Band Nummer: 1278
Autor: Danfeng Sun, M. Eng.
Ort: Hangzhou, China
ISBN: 978-3-18-527808-2
ISSN: 0178-9546
Erscheinungsjahr: 2022
Anzahl Seiten: 118
Anzahl Abbildungen: 67
Anzahl Tabellen: 12
Produktart: Buch (paperback, DINA5)

Produktbeschreibung

High requirements of industrial applications are challenging for WCSs, and advances in Artificial Intelligence (AI) bring a bright future in many fields. Hence, two directions from the aspects of WCS analysis and control are studied. Dependability assessment extracts dependability knowledge from given datasets, for which a series of AI methods are proposed, namely, deep autoencoder-based model, multi-task learning model, and device-level and system-level dependability assessment model. Spectral efficiency optimization aims to maximize spectral efficiency under certain conditions of quality of services. For this target, several AI models are proposed step by step for different conditions, namely, fully connected neural network, attention-based convolutional neural network, and generating precoder and power allocation neural network. Then, considering the resourceconstrained and distributed features, a distributed edge intelligence mechanism is proposed which can support AI applications on distributed edge devices. Evaluation on tasks of DA and SEO tasks indicates that the distributed edge intelligence has good executing efficiency.

Contents
Abbreviations vii
Zusammenfassung ix
Abstract xi
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Structure of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 General Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Related Work 6
2.1 Brief Introduction of Artificial Intelligence . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Milestones in Artificial Intelligence History . . . . . . . . . . . . . . . . 6
2.1.2 Artificial Intelligence Domains . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.3 Mainly Involved Machine Learning Techniques . . . . . . . . . . . . . . 10
2.2 AI Driven Wireless Communication Systems . . . . . . . . . . . . . . . . . . . 12
2.2.1 Subjects for AI Applications . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 AI to Improve Latency and Reliability . . . . . . . . . . . . . . . . . . . 13
2.2.3 AI to Increase Data Rate . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.4 AI for Connection Density . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.5 AI for Ensuring Safety and Security . . . . . . . . . . . . . . . . . . . . 17
2.3 AI for Focused Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1 Dependability Assessment from Analysis Phase . . . . . . . . . . . . . . 19
2.3.2 Spectral Efficiency Optimization from Control Phase . . . . . . . . . . . 20
2.3.3 Edge Intelligence in Wireless Communication Systems . . . . . . . . . . 20
2.4 Summary and Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3 Dependability Assessment 24
3.1 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.1.1 Formulation of Dependability Assessment . . . . . . . . . . . . . . . . . 25
3.1.2 Scope of Dependability Assessment Tasks . . . . . . . . . . . . . . . . . 26
3.1.3 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1.4 Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.5 Unsupervised Learning Process and Knowledge-based Process . . . . . . 31
3.2 Deep Autoencoder-based Model . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.1 Model Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.2 The Model Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.3 Deep Autoencoder for Dimension Reduction . . . . . . . . . . . . . . . 34
3.2.4 DBSCAN for Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . 34

ontents
3.2.5 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Multi-task Learning Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3.1 Problems of the Deep Autoencoder-based Model . . . . . . . . . . . . . 38
3.3.2 The Model Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3.3 Sequence to Sequence Learning . . . . . . . . . . . . . . . . . . . . . . 39
3.3.4 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4 Device Level and System Level Assessment . . . . . . . . . . . . . . . . . . . . 40
3.4.1 Further Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4.2 Device Level and System Level Assessment with the Distributed Edge
Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.3 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4 Spectral Efficiency Optimization 47
4.1 System Model and Optimization Problems . . . . . . . . . . . . . . . . . . . . . 47
4.2 Fully Connected Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.1 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.3 Attention-based Convolutional Neural Network . . . . . . . . . . . . . . . . . . 55
4.3.1 Problems of the Fully Connected Neural Network Model . . . . . . . . . 55
4.3.2 Network Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3.3 Applications of Attention Mechanism . . . . . . . . . . . . . . . . . . . 58
4.3.4 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4 Deep Learning-based Precoder and Power Allocation . . . . . . . . . . . . . . . 65
4.4.1 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.4.2 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5 Distributed Edge Intelligence for Dependability Assessment and Spectral
Efficiency Optimization 70
5.1 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.2 The Distributed Edge Intelligence Mechanism . . . . . . . . . . . . . . . . . . . 71
5.2.1 General System Structure . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.2.2 Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2.3 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.3 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3.1 Testbed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3.2 Distributed Edge Intelligence for the Dependability Assessment . . . . . 77
5.3.3 Distributed Edge Intelligence for the Spectral Efficiency Optimization . . 80
5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6 Conclusions and Future Work 83
6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Publication List 85
Bibliography 87

 

Keywords: Künstliche Intelligenz, Kabellose Kommunikationssysteme, industrielle Anwendungen, Zuverlässigkeitsbewertung, spektrale Effizienzoptimierung, Artificial Intelligence, Wireless Communication Systems, Industrial Applications, Dependability Assessment, Spectral Efficiency Optimization

48,00 € inkl. MwSt.
VDI-Mitgliedspreis:*
43,20 € inkl. MwSt.

* Der VDI-Mitgliedsrabatt gilt nur für Privatpersonen