Download the Full Ebook and Access More Features - ebooknice.

com

(Ebook) Connectivity and Edge Computing in IoT:

Customized Designs and AI-based Solutions by Gao,
Jie, Li, Mushu, Zhuang, Weihua ISBN 9783030887421,
3030887421
https://ebooknice.com/product/connectivity-and-edge-
computing-in-iot-customized-designs-and-ai-based-
solutions-36421838

OR CLICK HERE

DOWLOAD EBOOK

Download more ebook instantly today at https://ebooknice.com

 Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...

Start reading on any device today!

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason;

Viles, James ISBN 9781459699816, 9781743365571,
9781925268492, 1459699815, 1743365578, 1925268497
https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

ebooknice.com

(Ebook) Matematik 5000+ Kurs 2c Lärobok by Lena

Alfredsson, Hans Heikne, Sanna Bodemyr ISBN 9789127456600,
9127456609
https://ebooknice.com/product/matematik-5000-kurs-2c-larobok-23848312

ebooknice.com

(Ebook) SAT II Success MATH 1C and 2C 2002 (Peterson's SAT

II Success) by Peterson's ISBN 9780768906677, 0768906679

https://ebooknice.com/product/sat-ii-success-
math-1c-and-2c-2002-peterson-s-sat-ii-success-1722018

ebooknice.com

(Ebook) Master SAT II Math 1c and 2c 4th ed (Arco Master

the SAT Subject Test: Math Levels 1 & 2) by Arco ISBN
9780768923049, 0768923042
https://ebooknice.com/product/master-sat-ii-math-1c-and-2c-4th-ed-
arco-master-the-sat-subject-test-math-levels-1-2-2326094

ebooknice.com
(Ebook) Cambridge IGCSE and O Level History Workbook 2C -
Depth Study: the United States, 1919-41 2nd Edition by
Benjamin Harrison ISBN 9781398375147, 9781398375048,
1398375144, 1398375047
https://ebooknice.com/product/cambridge-igcse-and-o-level-history-
workbook-2c-depth-study-the-united-states-1919-41-2nd-edition-53538044

ebooknice.com

(Ebook) Intelligent Connectivity: AI, IoT, and 5G (IEEE

Press) by Abdulrahman Yarali ISBN 9781119685180,
1119685184
https://ebooknice.com/product/intelligent-connectivity-ai-iot-
and-5g-ieee-press-34907172

ebooknice.com

(Ebook) Building Enterprise IoT Solutions with Eclipse IoT

Technologies: An Open Source Approach to Edge Computing by
Frédéric Desbiens ISBN 9781484288818, 1484288815
https://ebooknice.com/product/building-enterprise-iot-solutions-with-
eclipse-iot-technologies-an-open-source-approach-to-edge-
computing-47394342
ebooknice.com

(Ebook) Building Enterprise IoT Solutions with Eclipse IoT

Technologies: An Open Source Approach to Edge Computing by
Frédéric Desbiens ISBN 9781484288825, 9781484288818,
1484288823, 1484288815
https://ebooknice.com/product/building-enterprise-iot-solutions-with-
eclipse-iot-technologies-an-open-source-approach-to-edge-
computing-47435426
ebooknice.com

(Ebook) Learning-based VANET Communication and Security

Techniques by Liang Xiao, Weihua Zhuang, Sheng Zhou,
Cailian Chen ISBN 9783030017309, 9783030017316,
3030017303, 3030017311
https://ebooknice.com/product/learning-based-vanet-communication-and-
security-techniques-7320634

ebooknice.com
Wireless Networks

Jie Gao
Mushu Li
Weihua Zhuang

Connectivity and
Edge Computing
in IoT: Customized
Designs and
AI-based Solutions
Wireless Networks

Series Editor
Xuemin Sherman Shen, University of Waterloo, Waterloo, ON, Canada
The purpose of Springer’s Wireless Networks book series is to establish the state
of the art and set the course for future research and development in wireless
communication networks. The scope of this series includes not only all aspects
of wireless networks (including cellular networks, WiFi, sensor networks, and
vehicular networks), but related areas such as cloud computing and big data.
The series serves as a central source of references for wireless networks research
and development. It aims to publish thorough and cohesive overviews on specific
topics in wireless networks, as well as works that are larger in scope than survey
articles and that contain more detailed background information. The series also
provides coverage of advanced and timely topics worthy of monographs, contributed
volumes, textbooks and handbooks.
** Indexing: Wireless Networks is indexed in EBSCO databases and DPLB **

More information about this series at http://www.springer.com/series/14180

Jie Gao • Mushu Li • Weihua Zhuang

Connectivity and Edge

Computing in IoT:
Customized Designs
and AI-based Solutions
Jie Gao Mushu Li
Department of Electrical and Computer Department of Electrical and Computer
Engineering Engineering
Marquette University University of Waterloo
Milwaukee, WI, USA Waterloo, ON, Canada

Weihua Zhuang
Department of Electrical and Computer
Engineering
University of Waterloo
Waterloo, ON, Canada

ISSN 2366-1186 ISSN 2366-1445 (electronic)

Wireless Networks
ISBN 978-3-030-88742-1 ISBN 978-3-030-88743-8 (eBook)
https://doi.org/10.1007/978-3-030-88743-8

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2021
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

The Internet of Things (IoT) is revolutionizing the world and impacting the daily
lives of billions of people. Supporting use cases for households, manufacturers,
transportation, agriculture, healthcare, and much more, IoT carries many potentials
and expectations for prospering human society. Technologically, we are at an early
stage of IoT development, aiming at connecting tens of billions of devices to make
homes, communities, factories, farms, and everywhere else smart and automated.
Tremendous efforts are necessary to advance IoT research and development.
Two cornerstones of IoT are data collection/exchange and data analysis. The
former demands connectivity solutions, while the latter requires computing solu-
tions. Due to the broad scope of IoT and the drastically different characteristics
and requirements of IoT use cases, no “one-size-fits-all” design can meet the
expectations of all use cases. Therefore, customizing connectivity or computing
solutions for specific use cases is challenging yet essential. There are many system
features and performance measures to consider in the customization, such as
connection link density, resource overhead, transmission and computation delay,
service reliability, energy efficiency, and device mobility, and making proper trade-
offs among them is critical.
Accounting for all performance metrics and making optimal trade-offs can yield
high complexity. Correspondingly, artificial intelligence (AI) solutions, such as
neural networks and reinforcement learning, can become useful. Powered by AI
methods, connectivity or computing solutions can learn from experience to handle
the complexity, assuming that sufficient data are available for training. Specifically,
AI can play various roles in IoT, including data traffic load prediction, access
control, and computation task scheduling, to name a few.
In this book, we focus on connectivity and edge computing in IoT and present
our designs for four representative IoT use cases, i.e., smart factory, rural IoT,
Internet of vehicles, and mobile virtual reality. We thoroughly review the existing
research in this field, including many works published in recent years. Then, through
innovative designs, we demonstrate the necessity and potential of customizing
solutions based on the use cases. In addition, we exploit AI methods to empower our
solutions. The four research works included in this book serve a collective objective:

v
vi Preface

enabling on-demand data collection and/or analysis for IoT use cases, especially in
resource-limited IoT systems. We hope that this book will inspire further research
on connectivity and edge computing in the field of IoT.

Milwaukee, WI, USA Jie Gao

Waterloo, ON, Canada Mushu Li
Waterloo, ON, Canada Weihua Zhuang
July 2021
Acknowledgements

The authors would like to thank Professor Xuemin (Sherman) Shen at the University
of Waterloo, the Editor of the Wireless Networks series, for his support in publishing
this book and his comments and suggestions on its technical content.
We would also like to thank Professor Lian Zhao at the Ryerson University for
her helpful discussions on the research presented in Chaps. 3 and 4.
In addition, we thank Dr. Xu Li, Professor Nan Cheng, and Conghao Zhou, who
participated in the research discussed in Chaps. 2, 4, and 5, respectively.
We appreciate valuable discussions with the members of the Broadband Com-
munications Research (BBCR) Lab at the University of Waterloo.
Special thanks to Susan Lagerstrom-Fife, Senior Editor at Springer New York,
and Shina Harshavardhan, Project Coordinator for Springer Nature for their help
during the preparation of this monograph.

vii
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 The Era of Internet of Things. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Connectivity in IoT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Edge Computing in IoT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 AI in IoT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Scope and Organization of This Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Industrial Internet of Things: Smart Factory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1 Industrial IoT Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Connectivity Requirements of Smart Factory . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.1 Application-Specific Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.2 Related Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.3 Potential Non-Link-Layer Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.4 Link-Layer Solutions: Recent Research Efforts . . . . . . . . . . . . . . . 16
2.3 Protocol Design for Smart Factory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1 Networking Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.2 Mini-Slot Based Carrier Sensing (MsCS) . . . . . . . . . . . . . . . . . . . . . 20
2.3.3 Synchronization Sensing (SyncCS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.4 Differentiated Assignment Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.5 Superimposed Mini-slot Assignment (SMsA) . . . . . . . . . . . . . . . . 26
2.3.6 Downlink Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4 Performance Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.4.1 Delay Performance with No Buffer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4.2 Delay Performance with Buffer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.3 Slot Idle Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.4 Impact of SyncCS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4.5 Impact of SMsA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5 Scheduling and AI-Assisted Protocol Parameter Selection . . . . . . . . . . . 34
2.5.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.5.2 The Considered Scheduling Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 36

ix
x Contents

2.5.3 Device Assignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.5.4 AI-Assisted Protocol Parameter Selection . . . . . . . . . . . . . . . . . . . . . 43
2.6 Numerical Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.6.1 Mini-Slot Delay with MsCS, SyncCS, and SMsA . . . . . . . . . . . . 46
2.6.2 Performance of the Device Assignment Algorithms . . . . . . . . . . 51
2.6.3 DNN-Assisted Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3 UAV-Assisted Edge Computing: Rural IoT Applications . . . . . . . . . . . . . . . . 63
3.1 Background on UAV-Assisted Edge Computing . . . . . . . . . . . . . . . . . . . . . . 63
3.2 Connectivity Requirements of UAV-Assisted MEC for Rural IoT . . . 65
3.2.1 Network Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2.2 State-of-the-Art Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3 Multi-Resource Allocation for UAV-Assisted Edge Computing . . . . . . 66
3.3.1 Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.3.2 Communication Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.3.3 Computing Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.3.4 Energy Consumption Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.3.5 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3.6 Optimization Algorithm for UAV-Assisted Edge
Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3.7 Proactive Trajectory Design Based on Spatial
Distribution Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.4 Numerical Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4 Collaborative Computing for Internet of Vehicles . . . . . . . . . . . . . . . . . . . . . . . 93
4.1 Background on Internet of Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.2 Connectivity Challenges for MEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.2.1 Server Selection for Computing Offloading . . . . . . . . . . . . . . . . . . 95
4.2.2 Service Migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.2.3 Cooperative Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.3 Computing Task Partition and Scheduling for Edge Computing . . . . . 97
4.3.1 Collaborative Edge Computing Framework . . . . . . . . . . . . . . . . . . . 97
4.3.2 Service Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.3.3 Service Failure Penalty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.3.4 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.3.5 Task Partition and Scheduling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.4 AI-Assisted Collaborative Computing Approach . . . . . . . . . . . . . . . . . . . . . 108
4.5 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.5.1 Task Partition and Scheduling Algorithm . . . . . . . . . . . . . . . . . . . . . 113
4.5.2 AI-Based Collaborative Computing Approach . . . . . . . . . . . . . . . . 114
4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Contents xi

5 Edge-Assisted Mobile VR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

5.1 Background on Mobile Virtual Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.2 Caching and Computing Requirements of Mobile VR . . . . . . . . . . . . . . . . 124
5.2.1 Mobile VR Video Formats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.2.2 Edge Caching for Mobile VR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.2.3 Edge Computing for Mobile VR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.3 Mobile VR Video Caching and Delivery Model . . . . . . . . . . . . . . . . . . . . . . 127
5.3.1 Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.3.2 Content Distribution Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.3.3 Content Popularity Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.3.4 Research Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.4 Content Caching for Mobile VR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.4.1 Adaptive Field-of-View Video Chunks . . . . . . . . . . . . . . . . . . . . . . . . 132
5.4.2 Content Placement on an Edge Cache . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.4.3 Placement Scheme for Video Chunks in a VS . . . . . . . . . . . . . . . . 139
5.4.4 Placement Scheme for Video Chunks of Multiple VSs . . . . . . . 142
5.4.5 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.5 AI-Assisted Mobile VR Video Delivery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
5.5.1 Content Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
5.5.2 Intelligent Content Distribution Framework. . . . . . . . . . . . . . . . . . . 150
5.5.3 WI-based Delivery Scheduling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.5.4 Reinforcement Learning Assisted Content Distribution . . . . . . 153
5.5.5 Neural Network Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
5.5.6 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
6.1 Summary of the Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
6.2 Discussion of Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Acronyms

3GPP Third generation partnership project

5G Fifth generation
5G NR 5G new radio
AC Actor-critic
AD Access delay
AD-F Access delay counted in frames
ADMM Alternating direction method of multipliers
AI Artificial intelligence
AP Access point
BFoV Base field-of-view
BS Base station
CNN Convolutional neural network
CSMA Carrier-sense multiple access
DCF Distributed coordination function
DDPG Deep deterministic policy gradient
DNN Deep neural network
DQN Deep Q network
DRL Deep reinforcement learning
EDT Early data transmission
EFoV Extended field-of-view
eMBB Enhanced mobile broadband
ET Enhanced tile
FoV Field-of-view
HMD Head-mounted device
HP High priority
IIoT Industrial Internet of Things
IoT Internet of Things
IoV Internet of Vehicles
IT Information technology
LoRa Long range
LP Low priority

xiii
xiv Acronyms

LPWA Low-power wide-area

LSTM Long short-term memory
LTE Long-term evolution
LTE-M Long-term evolution for machine-type communications
M2M Machine-to-machine
MAC Medium access control
MDP Markov decision process
mMTC Massive machine-type communications
mmWave Millimeter-wave
MsCS Mini-slot based carrier sensing
MSE Mean squared error
MTC Machine-type communication
NB-IoT Narrowband IoT
NOMA Non-orthogonal multiple access
QoE Quality of experience
QoS Quality of service
RACH Random access channel
RAW Restricted access window
RMAB Restless multi-armed bandit
RP Regular priority
RSU Roadside unit
SCA Successive convex approximation
SOC Second order cone
SMsA Superimposed mini-slot assignment
SyncCS Synchronization carrier sensing
TDMA Time-division multiple access
TPSA Task partition and scheduling algorithm
TTI Transmission time interval
UAV Unmanned aerial vehicle
URLLC Ultra-reliable low-latency communications
V2I Vehicle-to-infrastructure
V2X Vehicle-to-everything
VR Virtual reality
VS Video segment
WI Whittle index
WLAN Wireless local area network
Chapter 1
Introduction

In this chapter, we first provide an overview of the Internet of Things from the
perspectives of connected devices, use cases, deployment efforts, and technical
advancement. Then, connectivity and edge computing in IoT are introduced,
respectively, focusing on the requirements, available options, and challenges. The
role of artificial intelligence in IoT and challenges in developing AI-based solutions
are also discussed. Last, we present the scope and organization of this book.

1.1 The Era of Internet of Things

We are entering the era of the Internet of Things (IoT). Targeting to connect
billions of devices, such as wearables, appliances, and industrial actuators, and a
variety of systems, such as sensor networks, transportation management centers,
and power grids, IoT has become a major driver worldwide for innovations in
both business and technology development. The global IoT market size in 2020
is estimated to be approximately 309 billions in USD, and the forecast for 2021
and 2028 is 831 billions and 1855 billions, respectively, with an annual growth
rate of 25.4% between 2021 and 2028 [1]. Meanwhile, the number of networked
devices is expected to increase from around 20 billions in 2020 to almost 30 billions
in 2023, with almost 15 billion machine-to-machine (M2M) connections in 2023
[2]. Moreover, it is predicted that platforms connecting devices, cloud servers, and
application providers will harvest comparable revenue from emerging IoT use cases
and from traditional information technology (IT) use cases by 2023 [3].
IoT is a broad concept that covers a wide range of use cases. In manufacturing
industries, IoT solutions can improve asset management, optimize supply chains,
and enable factory automation [4]. In agriculture, IoT platforms can facilitate plant

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 1

J. Gao et al., Connectivity and Edge Computing in IoT: Customized Designs and
AI-based Solutions, Wireless Networks,
https://doi.org/10.1007/978-3-030-88743-8_1
2 1 Introduction

status monitoring, and pest and disease control [5]. In urban management, IoT
techniques can enable smart cities by integrating smart street lighting, intelligent
traffic control, fire and pollution detection, etc., to promote safe, comfortable,
and energy-efficient living conditions [6]. In healthcare, IoT applications can
support remote in-home health monitoring for proactive and preventive diagnosis
interventions [7]. In the airline business, IoT platforms can reduce fuel costs and
service disruption and thereby improve customer experience [8]. Other promising
IoT applications include crude oil production, wildfire detection, search and rescue,
smart campus, augmented shopping, and so on [9–13].
Many countries and regions have started IoT programs or pilot projects. For
example, the IoT European Large-Scale Pilots Programme has been promoting
partnerships across Europe since 2016 and conducting various IoT projects with
a total budget of e100 millions, including ACTIVAGE (for elderly smart living),
AUTOPILOT (for automated driving), IoF2020 (for the Internet of food and farm)
[14]. In the United States, New York City published its IoT strategy in March
2021 with an objective to create an IoT ecosystem for consumer, industry, and
government use cases [15], while other cities, such as Las Vegas, are on course to
become smart cities [16]. In China, the number of licensed IoT connections has
reached 600 millions by 2018, and a major focus of future IoT development is
intelligent manufacturing [17]. In addition to the above programs or pilot projects,
many industry leaders have invested in and developed IoT platforms, examples of
which include Amazon Web Services, Microsoft Azure IoT Platform, IBM Watson
IoT Platform, and Siemens MindSphere [18].
Besides various investment from governments and industries, technology
advancement in device hardware, software, communications, cloud/edge
computing, artificial intelligence (AI), etc., have been propelling the development
and deployment of IoT. Improvement in hardware enables the production of
IoT devices with smaller sizes and lower costs [19]. Improvements in software
allow IoT devices and platforms to become more secure, reliable, and energy-
efficient [20, 21]. Advancement in communication technologies enables a massive
scale of connections required for realizing IoT as well as new communication
paradigms, such as M2M communications [22, 23]. Modern cloud and edge
computing technologies provide versatile paradigms of data processing for IoT
applications, allowing on-demand computing service provisioning through task
offloading [24]. Lastly, advances in AI techniques render intelligent and automated
connectivity and computing solutions in IoT [25].
This book focuses on the connectivity and computing aspects of IoT, with a
particular focus on use case-specific designs and AI-based solutions. The rest of
this chapter will discuss the basics of connectivity, edge computing, and the role of
AI in IoT.
1.2 Connectivity in IoT 3

1.2 Connectivity in IoT

Connectivity is the foundation of IoT as it enables data collection from or exchange

among networked IoT devices. Different network topology, connectivity require-
ments, and connectivity options may apply in IoT, depending on the application.
Regarding network operation, IoT applications can be implemented in a dis-
tributed, a decentralized, or a centralized manner. Examples of distributed IoT
applications include distributed sensing and communication in autonomous driv-
ing [26] and plant monitoring for predictive maintenance in manufacturing [27],
which require a low response time and need to process collected data locally.
Examples of decentralized IoT applications include localization [28] and edge com-
puting [29], which leverage infrastructure and resources on network edge to serve
end users without relying on cloud servers. Additionally, many IoT applications
adopt a client-server mode and exploit centralized cloud computing platforms, such
as the enterprise platforms mentioned in Sect. 1.1. Examples of such applications
are metropolitan-area intelligent transportation system planning [30] and large-scale
supervisory control and data acquisition [31], which rely on extensive computing
and storage resources provided by data centers.
Regarding connectivity requirements, IoT applications may require high connec-
tion density, low communication delay, long communication range, high transmis-
sion rate, or combinations of those. In a smart city scenario, 30,000 connections
per square kilometer (km2 ) may be needed just for connecting household water,
electricity, and gas meters, which send messages with intervals between 30 min
and 24 h [32]. Such connections are delay-tolerant and usually have a short range,
e.g., 15 meters (m). In a factory automation setting, process state monitoring may
involve 10,000 devices per km2 [33]. Such connections span factory plants with
a typical size around 300 m × 300 m × 50 m, and the delay tolerance is on the
level of 50 milliseconds (ms). In internet of vehicles (IoV), a vehicle may need to
simultaneously communicate with hundreds of other vehicles [34]. The connections
for such communications can be transient, and the delay tolerance can be very strict,
e.g., 10 ms for road safety applications.
Regarding connectivity options, various wireless communication standards and
techniques are available for IoT. The three use cases of the fifth generation (5G)
cellular networks, i.e., enhanced mobile broadband (eMBB), ultra-reliable low-
latency communications (URLLC), and massive machine-type communications
(mMTC), aim at providing support for various IoT applications [35]. Meanwhile,
802.11ax, or Wi-Fi 6, has enhanced support for IoT and is suitable for smart home
applications [36]. In addition, a few low-power wide-area (LPWA) technologies
and standards, such as Long-Term Evolution for Machine-Type Communications
(LTE-M), Long Range (LoRa), Narrowband IoT (NB-IoT), support cost-effective
long-range communications and are suitable for applications such as smart logistics
and environment or wild-life monitoring [37]. In the future, IoT devices may also
be connected via satellites.
4 1 Introduction

Given the varieties of IoT applications and their connectivity requirements,

finding optimal connectivity solutions is challenging, and such challenge is aggran-
dized when considering network heterogeneity, device mobility, network resource
limitations, cost-effectiveness, and scalability. As a result, despite various potential
options as mentioned above, customized designs are necessary for providing the best
support to specific applications due to their unique characteristics and requirements.
In Chap. 2, we will customize a connectivity solution for industrial IoT and
demonstrate the potential of such customized designs for connecting IoT devices.
In Chaps. 3 and 4, we will present connectivity solutions related to computing task
offloading and result delivery in edge computing.

1.3 Edge Computing in IoT

Most IoT applications require not only data collection or exchange but also data
analysis. As a result, they demand a computing paradigm and related resources. The
data processing may happen on end user devices (such as sensors or vehicles), edge
facilities (such as local network controllers), or cloud computing servers (such as
Amazon Elastic Compute Cloud).
On-device processing is feasible for devices such as smartphones and vehicles,
which have the hardware, software, and other resources for on-board comput-
ing [38]. Meanwhile, a significant portion of IoT devices, such as sensors and
parking meters, are low-cost devices with limited processing power, storage, or
battery [39]. With no or minimum on-device processing capability, such devices
may resort to cloud computing and leverage resources in a cloud for data process-
ing [40]. The cloud computing paradigm enables a variety of IoT applications and
is especially suitable for applications running in a client-server mode. However,
cloud computing requires devices to upload the data for processing to a cloud server,
which can cause excessive traffic loads for the IoT networks when a massive number
of devices rely on cloud computing. In addition, the round-trip communication,
i.e., data uploading and computing result delivery, can cause a large delay that is
unacceptable for applications such as autonomous driving and industrial robot arm
control [41]. To reduce network traffic load and delay, edge computing has emerged
as a solution, in which computing resources are deployed outside of the cloud and
close to end users on network edge [42]. Such a computing paradigm is known as
mobile edge computing or multi-access edge computing (MEC).
With the advent of edge computing, applications that require low-delay com-
puting can leverage computing servers on the network edge [43]. This creates
new opportunities for both IoT service providers and network operators. In smart
healthcare, data collected by smartphones or wearable devices can be processed
at an edge server for health monitoring applications such as gait analysis and fall
risk assessment [44]. In smart cities, videos captured by cameras can be processed
at edge servers for surveillance and event recognition [45, 46]. In autonomous
driving, vehicles can upload data collected by cameras, radars, and other sensors
1.4 AI in IoT 5

to edge servers and enhance road safety via data analysis such as object recognition
and tracking. In addition, many applications in various domains that leverage edge
computing are emerging [47].
On the other hand, edge computing renders IoT networks more complex. New
challenges arise, which often involve the synergy of computing and connectivity.
For example, edge computing servers can be deployed at the access points (APs) of
femtocells (e.g., home networks), small cells, and macrocells, and each deployment
option has its own pros and cons [48]. In addition, the joint scheduling of
transmission and computing tasks becomes critical for supporting applications with
stringent delay requirements [49]. In highly dynamic networks such as vehicular
networks, computing service migration or collaborative computing can be necessary
for handling device mobility [50]. In Chaps. 3–5, we present edge computing
solutions in representative IoT scenarios, such as IoV, and discuss various issues
related to edge computing, such as task scheduling, content caching, collaborative
computing, and computing result delivery.

1.4 AI in IoT

The world has witnessed a rapid advancement of AI in the past decade, with
many successful real-world applications, especially in the field of natural language
processing and computer vision [51]. Such success inspires the investigation on
potential applications of AI in IoT, and many ideas have emerged for various use
cases, such as mining, healthcare, and transportation [52, 53].
Incorporating of AI in IoT is natural. First, involving a massive number of
devices, diverse applications, and spatiotemporally-variant service demands, IoT
networks are complex and dynamic. AI potentially offers a viable alternative
approach to managing IoT networks with the desired scalability and adaptability,
while satisfying diverse and often stringent application requirements. Second, the
effectiveness of AI relies on abundant data, e.g., for training neural networks, while
a massive number of IoT devices can generate or provide a massive amount of data
to fuel AI. Last, AI methods are suitable for data analysis in many IoT applications,
such as health monitoring and fault pattern identification in smart grids [54].
AI can play a multifarious role in IoT, in terms of both the connectivity and the
edge computing. Specifically, AI can be used for network traffic load prediction
to facilitate IoT network planning [55]. AI can also be adopted in medium access
control (MAC) to enhance IoT network throughput or fairness [56]. In addition,
AI can be applied to handle computing task scheduling [57], offloading [58],
and migration [59] for effective edge computing with minimum computing delay,
balanced computing load distribution, or adaptivity to network dynamics.
Despite a tremendous potential of AI in empowering various IoT applications,
many challenges exist in AI-based solutions for IoT. Specifically, choosing appro-
priate AI methods for considered IoT applications, while taking practicality into
account, is essential yet challenging. Moreover, AI functionality deployment, com-
6 1 Introduction

munication overhead, data processing delay, and scalability of AI-based solutions,

among other possible issues, all need to be accounted for. In Chaps. 2, 4, and 5,
we develop AI-based connectivity and edge computing solutions in representative
IoT scenarios, including learning assisted scheduling, collaborative computing, and
content distribution. These solutions demonstrate the potentials and advantages of
incorporating AI into various IoT applications.

1.5 Scope and Organization of This Book

In this book, we focus on the connectivity and edge computing aspects of IoT. We
develop customized designs and AI-based solutions for connectivity and/or edge
computing in representative IoT use cases, including smart factory, rural IoT, IoV,
and mobile virtual reality (VR).
In Chap. 2, we investigate MAC for an industrial IoT network. Considering a
local area network with high device density, short packets, and stringent delay and
reliability requirements, we tailor a MAC protocol for smart factory applications
and design a neural network to assist the scheduling of transmission opportunities
for industrial IoT devices.
In Chap. 3, we investigate unmanned aerial vehicle (UAV) assisted edge com-
puting for rural IoT applications such as in smart agriculture or forest monitoring.
Using a UAV to provide connectivity and computing service to IoT devices, we
develop a solution to jointly optimize the connectivity, through determining the UAV
trajectory and device transmit power, and the edge computing, through properly
allocating computing load between the UAV and the devices.
In Chap. 4, we investigate edge computing for delay-sensitive applications
in IoV to improve the safety or driving experience of drivers. To address the
challenge of high vehicle mobility, we adopt collaborative edge computing to reduce
computing delay and improve computing service reliability for vehicles and develop
a deep reinforcement learning assisted approach to find the optimal computing task
offloading and computing result delivery policy.
In Chap. 5, we investigate edge-assisted content caching and distribution for
mobile VR video streaming, which requires edge computing to render some VR
videos. To improve the viewer’s quality of experience (QoE), we design a scheme to
cache video content and reduce frame missing in VR video streaming, and develop
a deep reinforcement learning based scheme for scheduling VR content delivery to
viewers.
In Chap. 6, we conclude this book and briefly discuss further research directions
in connectivity and edge computing in IoT.
References 7

References

1. Internet of Things (IoT) market size, share & covid-19 impact analysis, by component
(platform, solution & services), by end use industry (BFSI, retail, government, healthcare,
manufacturing, agriculture, sustainable energy, transportation, IT & telecom, others), and
regional forecast, 2021-2028. Tech. Rep. Report ID: FBI100307, Fortune Business Insights
2. Cisco annual Internet report (2018–2023), white paper. Tech. rep., Cisco
3. Dahlqvist, F., Patel, M., Rajko, A., Shulman, J.: Growing opportunities in the Internet of
Things. Tech. rep., McKinsey (2019)
4. Yang, C., Shen, W., Wang, X.: The Internet of Things in manufacturing: key issues and
potential applications. IEEE Syst. Man Cybern. Mag. 4(1), 6–15 (2018)
5. Elijah, O., Rahman, T.A., Orikumhi, I., Leow, C.Y., Hindia, M.N.: An overview of Internet of
Things (IoT) and data analytics in agriculture: benefits and challenges. IEEE Internet Things
J. 5(5), 3758–3773 (2018)
6. Du, R., Santi, P., Xiao, M., Vasilakos, A.V., Fischione, C.: The sensable city: a survey on the
deployment and management for smart city monitoring. IEEE Commun. Surv. Tut. 21(2),
1533–1560 (2019)
7. Philip, N.Y., Rodrigues, J.J.P.C., Wang, H., Fong, S.J., Chen, J.: Internet of Things for in-home
health monitoring systems: current advances, challenges and future directions. IEEE J. Sel.
Areas Commun. 39(2), 300–310 (2021)
8. Lavoie-Tremblay, K., Gautam, S., Levine, G.: Connecting the dots on IoT for the industrial
world. IEEE Internet Things Mag. 1(1), 24–26 (2018)
9. Duan, Q., Sun, D., Li, G., Yang, G., Yan, W.W.: IoT-enabled service for crude-oil production
systems against unpredictable disturbance. IEEE Trans. Services Comput. 13(4), 759–768
(2020)
10. Bushnaq, O.M., Chaaban, A., Al-Naffouri, T.Y.: The role of UAV-IoT networks in future
wildfire detection. IEEE Internet Things J. (2021). https://doi.org/10.1109/JIOT.2021.3077593
11. Bianco, G.M., Giuliano, R., Marrocco, G., Mazzenga, F., Mejia-Aguilar, A.: LoRa system for
search and rescue: path-loss models and procedures in mountain scenarios. IEEE Internet
Things J. 8(3), 1985–1999 (2021)
12. Sutjarittham, T., Habibi Gharakheili, H., Kanhere, S.S., Sivaraman, V.: Experiences with IoT
and AI in a smart campus for optimizing classroom usage. IEEE Internet Things J. 6(5), 7595–
7607 (2019)
13. Hormann, L.B., Putz, V., Rudic, B., Kastl, C., Klinglmayr, J., Pournaras, E.: Augmented
shopping experience for sustainable consumption using the Internet of Thing. IEEE Internet
Things Mag. 2(3), 46–51 (2019)
14. IoT European large-scale pilots programme - large scale pilots projects. IoT European Large-
Scale Pilots Programme (2018). https://european-iot-pilots.eu/wp-content/uploads/2018/03/
220315_SD_IoT_Brochure_A4_LowRes_final-1.pdf
15. The New York City Internet of Things strategy. version 1.26.0402 (2021). https://www1.nyc.
gov/assets/cto/downloads/iot-strategy/nyc_iot_strategy.pdf
16. Dickens, C., Boynton, P., Rhee, S.: Principles for designed-in security and privacy for smart
cities. In: Proceedings of the 4th Workshop Int. Sci. Smart City Operations Platforms Eng.
(SCOPE), pp. 25–29 (2019)
17. Giaffreda, R.: IoT in China: What does the future hold? IEEE Internet Things Mag. 2(3), 52–53
(2019)
18. Hoffmann, J.B., Heimes, P., Senel, S.: IoT platforms for the Internet of Production. IEEE
Internet Things J. 6(3), 4098–4105 (2019)
19. Folea, S.C., Mois, G.D.: Lessons learned from the development of wireless environmental
sensors. IEEE Trans. Instrum. Meas. 69(6), 3470–3480 (2020)
20. Atlam, H.F., Wills, G.B.: IoT security, privacy, safety and ethics. In: Digital Twin Technologies
and Smart Cities, pp. 123–149. Springer, Cham (2020)
8 1 Introduction

21. Georgiou, K., Xavier-de Souza, S., Eder, K.: The IoT energy challenge: a software perspective.
IEEE Embedded Syst. Lett. 10(3), 53–56 (2018)
22. Gazis, V.: A survey of standards for machine-to-machine and the Internet of Things. IEEE
Commun. Surv. Tut. 19(1), 482–511 (2017)
23. Gao, J., Li, M., Zhao, L., Shen, X.: Contention intensity based distributed coordination for
V2V safety message broadcast. IEEE Trans. Veh. Technol. 67(12), 12,288–12,301 (2018)
24. Porambage, P., Okwuibe, J., Liyanage, M., Ylianttila, M., Taleb, T.: Survey on multi-access
edge computing for Internet of Things realization. IEEE Commun. Surv. Tut. 20(4), 2961–
2991 (2018)
25. Javaid, N., Sher, A., Nasir, H., Guizani, N.: Intelligence in IoT-based 5G networks: opportuni-
ties and challenges. IEEE Commun. Mag. 56(10), 94–100 (2018)
26. Philip, B.V., Alpcan, T., Jin, J., Palaniswami, M.: Distributed real-time IoT for autonomous
vehicles. IEEE Trans. Ind. Inf. 15(2), 1131–1140 (2019)
27. Liu, Y., Yu, W., Dillon, T.S., Rahayu, W., Li, M.: Empowering IoT predictive maintenance
solutions with AI: a distributed system for manufacturing plant-wide monitoring. In: IEEE
Transactions on Industrial Informatics (2021)
28. Kasmi, Z., Guerchali, N., Norrdine, A., Schiller, J.H.: Algorithms and position optimization
for a decentralized localization platform based on resource-constrained devices. IEEE Trans.
Mobile Comput. 18(8), 1731–1744 (2019)
29. Cicconetti, C., Conti, M., Passarella, A.: A decentralized framework for serverless edge
computing in the Internet of Things. IEEE Trans. Netw. Service Manag. 18(2), 2166–2180
(2021)
30. Lidkea, V.M., Muresan, R., Al-Dweik, A.: Convolutional neural network framework for
encrypted image classification in cloud-based ITS. IEEE Open J. Intell. Transp. Syst. 1, 35–50
(2020)
31. Sajid, A., Abbas, H., Saleem, K.: Cloud-assisted IoT-based scada systems security: a review
of the state of the art and future challenges. IEEE Access 4, 1375–1384 (2016)
32. Ericsson mobility report - on the pulse of the networked society. Tech. rep., Ericsson (2016)
33. Technical specification group services and system aspects; service requirements for the 5G
system; stage 1 (Release 16). Tech. Rep. TS 22.261, Version 16.14.0, 3GPP (2021)
34. Bi, Y., Zhou, H., Zhuang, W., Zhao, H.: Safety Message Broadcast in Vehicular Networks.
Springer, Cham (2017)
35. Li, J., Shi, W., Yang, P., Ye, Q., Shen, X.S., Li, X., Rao, J.: A hierarchical soft RAN slicing
framework for differentiated service provisioning. IEEE Wireless Commun. 27(6), 90–97
(2020)
36. Wi-Fi 6 industry impact report - transition or transformation? Tech. rep., FeibusTech
37. Raza, U., Kulkarni, P., Sooriyabandara, M.: Low power wide area networks: an overview. IEEE
Commun. Surv. Tut. 19(2), 855–873 (2017)
38. Rawassizadeh, R., Pierson, T.J., Peterson, R., Kotz, D.: NoCloud: exploring network discon-
nection through on-device data analysis. IEEE Pervasive Comput. 17(1), 64–74 (2018)
39. Ciuffoletti, A.: Low-cost IoT: a holistic approach. J. Sens. Actuator Netw. 7(2), 19 (2018)
40. Barcelo, M., Correa, A., Llorca, J., Tulino, A.M., Vicario, J.L., Morell, A.: IoT-cloud service
optimization in next generation smart environments. IEEE J. Sel. Areas Commun. 34(12),
4077–4090 (2016)
41. Zhang, B., Mor, N., Kolb, J., Chan, D.S., Lutz, K., Allman, E., Wawrzynek, J., Lee, E.,
Kubiatowicz, J.: The cloud is not enough: Saving IoT from the cloud. In: Proceedings of
the 7th USENIX Workshop Hot Topics Cloud Comput. (HotCloud 15). USENIX Association,
Santa Clara, CA (2015)
42. Shi, W., Cao, J., Zhang, Q., Li, Y., Xu, L.: Edge computing: vision and challenges. IEEE
Internet Things J. 3(5), 637–646 (2016)
43. Ashouri, M., Davidsson, P., Spalazzese, R.: Cloud, edge, or both? Towards decision support for
designing IoT applications. In: Proc. Fifth Int. Conf. Internet Things: Syst., Manage. Secur.,
pp. 155–162 (2018)
References 9

44. Baktir, A.C., Tunca, C., Ozgovde, A., Salur, G., Ersoy, C.: SDN-based multi-tier computing and
communication architecture for pervasive healthcare. IEEE Access 6, 56765–56781 (2018)
45. Long, C., Cao, Y., Jiang, T., Zhang, Q.: Edge computing framework for cooperative video
processing in multimedia IoT systems. IEEE Trans. Multimedia 20(5), 1126–1139 (2018)
46. Liu, S., Liu, L., Tang, J., Yu, B., Wang, Y., Shi, W.: Edge computing for autonomous driving:
opportunities and challenges. Proc. IEEE 107(8), 1697–1716 (2019)
47. Porambage, P., Okwuibe, J., Liyanage, M., Ylianttila, M., Taleb, T.: Survey on multi-access
edge computing for Internet of Things realization. IEEE Commun. Surv. Tut. 20(4), 2961–
2991 (2018)
48. Vallati, C., Virdis, A., Mingozzi, E., Stea, G.: Mobile-edge computing come home connecting
things in future smart homes using LTE device-to-device communications. IEEE Consum.
Electron. Mag. 5(4), 77–83 (2016)
49. He, H., Shan, H., Huang, A., Ye, Q., Zhuang, W.: Edge-aided computing and transmission
scheduling for LTE-U-enabled IoT. IEEE Trans. Wireless Commun. 19(12), 7881–7896 (2020)
50. Wang, S., Guo, Y., Zhang, N., Yang, P., Zhou, A., Shen, X.: Delay-aware microservice
coordination in mobile edge computing: a reinforcement learning approach. IEEE Trans.
Mobile Comput. 20(3), 939–951 (2021)
51. Wiriyathammabhum, P., Summers-Stay, D., Fermüller, C., Aloimonos, Y.: Computer vision
and natural language processing: recent approaches in multimedia and robotics. ACM Comput.
Surv. 49(4), 1–44 (2016)
52. Al-Turjman, F.: Artificial Intelligence in IoT. Springer, Cham (2019)
53. Manoharan, K.G., Nehru, J.A., Balasubramanian, S.: Artificial Intelligence and IoT: Smart
Convergence for Eco-friendly Topography, vol. 85. Springer Nature, Singapore (2021)
54. Bose, B.K.: Artificial intelligence techniques in smart grid and renewable energy systems–
some example applications. Proc. IEEE 105(11), 2262–2273 (2017)
55. Wang, W., Zhou, C., He, H., Wu, W., Zhuang, W., Shen, X.: Cellular traffic load prediction
with LSTM and Gaussian process regression. In: Proceedings of the 2020 IEEE International
Conference Communication (ICC), pp. 1–6 (2020)
56. Yu, Y., Wang, T., Liew, S.C.: Deep-reinforcement learning multiple access for heterogeneous
wireless networks. IEEE J. Sel. Areas Commun. 37(6), 1277–1290 (2019)
57. Zhou, C., Wu, W., He, H., Yang, P., Lyu, F., Cheng, N., Shen, X.: Deep reinforcement learning
for delay-oriented IoT task scheduling in SAGIN. IEEE Trans. Wireless Commun. 20(2), 911–
925 (2021)
58. Ye, Q., Shi, W., Qu, K., He, H., Zhuang, W., Shen, X.: Joint RAN slicing and computation
offloading for autonomous vehicular networks: a learning-assisted hierarchical approach. IEEE
Open J. Veh. Technol. 2, 272–288 (2021)
59. Yuan, Q., Li, J., Zhou, H., Lin, T., Luo, G., Shen, X.: A joint service migration and mobility
optimization approach for vehicular edge computing. IEEE Trans. Veh. Technol. 69(8), 9041–
9052 (2020)
Chapter 2
Industrial Internet of Things: Smart
Factory

In this chapter, we investigate the smart factory use case in the scenario of
industrial IoT, focusing on the connectivity aspect. First, through reviewing the
connectivity requirements and related standards, we illustrate the insufficiency of
existing techniques in meeting the expectations of smart factories and the necessity
of tailoring connectivity solutions. Then, we design a novel medium access control
protocol, which features grant-free distributed channel access, to support high
device density and low communication latency with low communication overhead.
We further propose a deep neural network-assisted centralized approach to configure
the protocol parameters and schedule transmission opportunities for all devices.
Combining the customized protocol and the AI-assisted scheduling, our design
demonstrates promising potentials for smart factories by simultaneously enabling
massive connections and millisecond-level delay for high priority devices.

2.1 Industrial IoT Networks

Industrial Internet of Things (IIoT) utilizes connected devices, including sensors,

actuators, and controllers to automate data collection and analysis for increasing
productivity, reducing energy consumption, and improving safety and reliability in
various industries, such as manufacturing, construction, warehouses, and oil rigs
and refineries [1, 2]. As one of the most promising technology domains, IIoT is
envisioned to reshape industries around the world, e.g., creating “factories of the
future”. As a result, IIoT related research and development are attracting widespread
attention, and the global IIoT market is expected to reach 263.4 billions in US
dollars by 2027 [3].

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 11

J. Gao et al., Connectivity and Edge Computing in IoT: Customized Designs and
AI-based Solutions, Wireless Networks,
https://doi.org/10.1007/978-3-030-88743-8_2
12 2 Industrial Internet of Things: Smart Factory

Industrial communication networks will play a crucial role in the upcoming

IIoT [4]. As in the general IoT, machine-type communication (MTC) is a primary
enabler of IIoT networks. Features of MTC that are recognized by the 3rd
Generation Partnership Project (3GPP) include the following [5]:
• Small packet transmission: MTC devices usually transmit and receive small
amounts of data, e.g., 1 kilobyte (KB) data size;
• Time controlled access: MTC devices may tolerate communicating in predefined
time intervals to avoid signaling overhead;
• Low mobility: MTC devices either remain at the same location or move
infrequently within a limited area;
• Monitoring: the network should be able to detect events such as the loss of
connectivity, change of location, and communication failure;
• Group-based features: the network should support the grouping of MTC devices
and the association of devices to groups.
Besides the above features mentioned by 3GPP, additional features of MTC have
been identified in the literature [6]:
• Uplink-dominated transmissions: The uplink traffic can be caused by a vast
number of sensors sending data to an AP;
• Low data rate: Typical data rate for MTC ranges from 100 kilobits per second
(kbps) to 10 megabits per second (mbps);
• Sporadic transmissions: Packet inter-arrival time at each device may range from
several milliseconds to several minutes [7];
• Low-complexity devices: MTC devices are usually cost-constrained and may not
support complex on-board processing.
Compared with the general IoT, IIoT has some unique characteristics. First,
connectivity in IIoT is usually structured, featuring centralized network manage-
ment [8]. Second, IIoT scenarios generally involve densely deployed devices in
a relatively limited area. For example, process monitoring in IIoT may involve
10, 000 devices per km2 [9]. Third, certain IIoT applications are mission-critical
and have extremely stringent quality of service (QoS) requirements. For example,
the communication latency tolerance for machine tool motion control can be
less than 0.5 ms [10]. The combination of the above characteristics poses a
significant challenge for supporting MTC in IIoT. Specifically, within a limited
geographical area, such as a factory, a communication network may need to support
a massive number of devices and, simultaneously, satisfy exceptionally strict QoS
requirements for some devices.
Existing standards, including LTE-M, NB-IoT, IEEE 802.11ah, and 5G new
radio (NR), are not sufficient for supporting MTC in IIoT networks. For example,
LTE-M and NB-IoT, both targeting low-power wide-area communications, are more
concerned with radio spectrum usage and power consumption than communication
delay. As for 5G NR, the delay threshold to support 106 connections per km2 is set to
10 s in the link-level simulations conducted by 3GPP, while a low packet arrival rate,
i.e., 1 packet per 2 h per device, is used [11]. More details on these standards will
2.2 Connectivity Requirements of Smart Factory 13

be given in Sect. 2.2.2. To address the challenge in simultaneously supporting high

device density and satisfying stringent QoS requirements, various solutions have
been proposed for different layers in the network protocol stack. At the physical
layer, utilizing spectrum resources beyond 30GHz or adopting nonorthogonal
multiple access (NOMA) can provide support for a high device density. However,
physical layer solutions have limitations in terms of transceiver hardware and
signal processing complexities, cost-effectiveness, and signaling overhead. At the
link layer, new MAC designs and enhancements have been proposed for cellular
networks and wireless local area networks, which we will discuss in detail in
Sect. 2.2.4.
While the existing studies provide important insights, further research on
customized MAC protocols is necessary for smart factories and, in particular, for
applications such as factory automation and process control. In the rest of this
chapter, we present the communication requirements of the smart factory use
case, review the related solutions, tailor a MAC design for smart factories, and
demonstrate the performance of our design.

2.2 Connectivity Requirements of Smart Factory

In a smart factory, various devices are connected in order to collect, share, and ana-
lyze data for improving the productivity and safety of manufacturing while reducing
costs. The smart factory use case involves many applications, most of which require
a fast and reliable communication network. In this section, we introduce application-
specific requirements, review existing standards, and summarize recent research
efforts on enabling smart factories.

2.2.1 Application-Specific Requirements

Different applications have different connectivity requirements. Table 2.1 shows

some representative applications and their requirements [10]. In the table, “cycle
time” refers to the transmission interval in periodic communication, which is usually
larger than the acceptable communication delay. In addition to the applications in
this table, more information on application-specific connectivity requirements can
be found in Annex E of 3GPP TR 22.804 [12].
Table 2.1 shows that the smart factory use case features a wide range of
applications and encompasses a variety of devices such as mobile robots, milling
machines, and automated guided vehicles. Accordingly, an IIoT network will need
to simultaneously support different applications and their performance require-
ments. For example, cooperative motion control for mobile robots has a cycle time
of 1 ms and thus a stringent communication delay tolerance, while the number of
mobile robots in a network is limited. By contrast, process automation or monitoring
14 2 Industrial Internet of Things: Smart Factory

Table 2.1 Representative applications and connectivity requirements [10]

Applications Cycle time Typical area Number of devices
Printing machine <2 ms 100 m × 100 m × 30 m >100
Motion Control Machine tool <0.5 ms 15 m × 15 m × 3 m ∼20
Packaging <1 ms 10 m × 5 m × 3 m ∼50
machine
Cooperative 1 ms <1 km2 100
Mobile robots
motion control
Video-operated 10–100 ms <1 km2 100
remote control
Mobile control Assembly robots/ 4–8 ms 10 m × 10 m 4
panels with safety milling machines
functions Mobile cranes 12 ms 40 m × 60 m 2
Process automation/monitoring >50 ms 104 devices per km2

has a cycle time of 50 ms or more and thus a relatively less stringent communication
delay tolerance, while the number of devices to be connected is large. Therefore, an
industrial network must achieve the following targets:
• Connecting a large number of devices with assorted types;
• Satisfying different performance requirements, some of which can be highly
stringent, for different devices and applications.
From the above two targets, we can see that future industrial networks need to
simultaneously support URLLC and mMTC. Next, we review existing standards
and recent research efforts related to IIoT.

2.2.2 Related Standards

In recent years, new standards for supporting MTC have been emerging. Existing
standards for MTC can be categorized into two groups: cellular-based and non-
cellular-based standards. The most representative examples of the former are NB-
IoT, LTE-M, and 5G NR, while the most representative examples of the latter are
IEEE 802.11ah and proprietary standards such as LoRa. Among these protocols,
IEEE 802.11ah is a wireless local area network (WLAN) standard, and the others
are wide-area network standards.
NB-IoT debuted in 3GPP Release 13 and can support up to 106 connections
per km2 [13] or 5 × 104 connections per cell [14] with a low data rate (160 kbps
or less) using a 180 kHz channel bandwidth. Moreover, NB-IoT devices can save
device battery power by remaining in low-power mode in-between transmissions.
However, the cost for energy efficiency is a large delay. The typical delay of NB-
IoT is larger than 1 second (s) and sometimes as large as 10 s, which is unacceptable
for many IIoT applications. Compared with NB-IoT, the delay performance of LTE-
2.2 Connectivity Requirements of Smart Factory 15

M is better but still far from meeting the expectation of IIoT. The LTE-M uses a
much larger bandwidth, i.e., 5 MHz as opposed to 180 kHz, and supports a much
higher data rate, up to 7 Mbps. The reported connection density of LTE-M varies
in the range from 104 [14] to 8.5 × 104 connections per cell [15].1 While the delay
performance of LTE-M is better than that of NB-IoT, it is still larger than 100 ms
under low network load and can easily increase to 1 s or more under high network
load [16].
5G NR can support either massive connections or millisecond-level delay.
However, 3GPP Release 15 lists “Critical Communications (CC) and URLLC”
and “massive Internet of Things (mIoT)” as two separate service aspects [17].
Consequently, link-level simulations conducted by 3GPP demonstrate that 5G NR
can satisfy the URLLC performance requirement of 1 ms delay or the mMTC
performance requirement of 106 connections per km2 , but not at the same time.
Specifically, the delay threshold for NR to support 106 connections per km2 is set
to 10 s in the above evaluation, while the packet arrival rate is 1 packet per 2 h
per device [11], which is much lower than the cycle times in Table 2.1. Note that
3GPP Release 16 improves support for IoT through enhancements in scheduling
and network reference time synchronization. Nevertheless, more improvement is
necessary for simultaneously achieving high density and low latency [18].
Using a bandwidth from 125 kHz to 500 kHz, the propriety standard LoRa
could achieve a theoretical transmission range of 10 km. However, the probability
of successful transmission at the first attempt is below 0.2 when 1000 devices are
connected [19]. Such reliability performance can be unacceptable for many smart
factory applications. Moreover, the delay of LoRa is no less than 1 s if 100 or more
devices are connected and can easily surpass 10 s with further increased connection
density [20].2 The high delay and low reliability limit the use of LoRa in the smart
factory use case.
IEEE 802.11ah can support a less than 10 ms delay but only under a low load
condition [21]. For example, 802.11ah is suitable for infrequent transmission with a
packet inter-arrival duration of no less than 30 s. When 500 devices are connected
to the AP, the delay becomes approximately 300 ms even under the relatively long
packet inter-arrival time of 180 s [22]. Recent works propose improvements of
IEEE 802.11ah, yet the improvement is mostly seen in the throughput rather than
in the delay [23, 24]. The lack of guarantee on a millisecond-level delay limits the
suitability of IEEE 802.11ah for smart factories.
In summary, the insufficiency of existing standards for supporting MTC in dense
industrial networks is clear. None of the existing standards can simultaneously
support massive connections and guarantee a millisecond-level access delay, which

1 The upper limit is calculated based on data in [15] as follows: 357,000 devices per 1.08 MHz

multiplied by 5 MHz and then divided by 21 cells.

2 In practice, connection density can be higher than device density due to dual-connectivity or

multi-connectivity. We use “connection density” and “device density” interchangeably since multi-
connectivity is not a focus of this book.
16 2 Industrial Internet of Things: Smart Factory

is needed for the smart factory use case. Therefore, there is an urgent need for
developing and standardizing solutions that can meet the needs of the smart factory
use case.

2.2.3 Potential Non-Link-Layer Solutions

To address the challenge of supporting high device density and stringent QoS
requirements, various solutions have been proposed.
In cellular networks, network densification and network slicing allow networks
to support high user density and satisfy stringent QoS requirements. In addition
to supporting high connection density, network densification can reduce link access
delay [25]. However, there is a limitation on the density of APs due to the increasing
cost, and interference and the diminishing performance gain as the network densi-
fies [26]. Network slicing enables flexible service provision for coexisting services
with different QoS requirements [27]. As a result, it can potentially contribute to
simultaneously supporting mMTC and URLLC in IIoT scenarios. However, the
complexity for network slicing can be high due to the need for frequent resource
reservation and orchestration [28]. For industrial networks, especially medium or
small industrial networks, network densificatoin or network slicing may not be an
ideal solution.
Emerging physical-layer techniques, such as millimeter-wave (mmWave) com-
munication and NOMA, could also contribute to supporting the smart factory use
case. Extending the available radio spectrum to the mmWave range could help
support high connection density. However, equipping every sensor, actuator, and
other devices with hardware for mmWave communications can yield a high cost,
which may hinder the practical deployment of smart factories. Some research works
have proposed NOMA solutions for mMTC, such as compressed sensing based
multi-user detection [6], coded tandem spreading [29], and block sparsity and block
precoding [30]. Such solutions usually require either advanced signal processing,
which increases algorithm and hardware complexity, or availability of transmitter
side information at receivers, which results in signaling overhead [6, 31].

2.2.4 Link-Layer Solutions: Recent Research Efforts

Different from physical-layer solutions, link-layer solutions can be flexible as they

can be implemented through software. In addition, link-layer solutions can be cost-
effective and customized based on the application QoS requirements. As a result,
link-layer solutions have tremendous potential for supporting the smart factory use
case. Existing studies on link-layer solutions can be categorized into solutions for
cellular networks, solutions for WLANs, and hybrid solutions.
2.2 Connectivity Requirements of Smart Factory 17

For cellular networks, the bottleneck is the contention-based random access

channel (RACH) procedure for connection setup. Specifically, a network can be
congested when a massive number of devices try to establish connections around
the same time [32]. In 3GPP release 15, the design of early data transmission (EDT)
replaces a standard four-step RACH procedure with a two-step procedure [17].
In existing research works, prioritization and grouping have been a focus of
refining the RACH procedure, and different ideas have appeared. Devices can be
grouped, e.g., based on their delay requirements, to limit the collision probability.
Groups contend with each other to make access attempts, while either one [33] or
multiple groups [34] can be active at a given instant. Alternatively, devices can be
grouped and later redistributed into different groups after encountering an initial
collision [35]. Access class barring [36], extended access class barring [37], and
their derivatives [38–40], as distributed coordinate mechanisms, have also gained
popularity and attracted much attention in the literature. After the connection
setup stage, scalable transmission time intervals (TTI) [41, 42], and preemption
of scheduled low-priority transmissions [43] for high-priority devices can be
used for reducing access delay. Overall, most solutions represent refinements of
existing protocol designs, e.g., refining the RACH procedure. These solutions often
make a trade-off between different performance metrics, e.g., delay and collision
probability, while simultaneously improving performance over several metrics is
necessary for the smart factory use case. Moreover, the above solutions are grant-
based, which causes unnecessary overhead and delay, while grant-free access is
preferred [44].
For WLANs, improvements over 802.11ah are the focus of many works on
MAC design. The key mechanism in 802.11ah is the restricted access window
(RAW), which groups devices and allows channel access for different groups
in different time durations. Existing MAC solutions for LAN focus on refining
RAW. For example, one category of works optimizes the RAW window size based
on the group size [45, 46], dynamically changing the window size according to
failed transmission attempts [47], or allocating RAW slots based on group QoS
requirements [48]. Another category of works improves the grouping method for
RAW according to geographical device distribution [49], traffic volume [50], data
rate [23], or potential hidden terminal relationship among nodes [51]. Besides the
above two categories, other efforts to improve RAW include periodical RAW for
periodic traffic scheduling [52]. Compared with cellular-based solutions, WLAN-
based solutions have the advantage of grant-free access and lower overhead, while
the disadvantage is lower reliability due to inevitable collisions in data transmission.
Consequently, the connection density that can be supported by WLAN-based
solutions is generally smaller than that can be supported by cellular-based solutions,
while the collision probability and delay can be high in the case of a large number
of devices [53].
Moreover, there are hybrid solutions that combine or switch between grant-based
access in cellular and grant-free contention-based random access in WLAN. For
example, a MAC solution that splits data traffic volume between the two radio access
technologies is proposed in [54] for heterogeneous networks in which both cellular
18 2 Industrial Internet of Things: Smart Factory

and WLAN APs are available. Another example is switching between time-division
multiple access (TDMA) and distributed coordination function (DCF), depending
on the data traffic load [55]. Additionally, hybrid MAC that implements carrier-
sense multiple access (CSMA) and TDMA in different stages of data transmission
exists in the literature [56]. These solutions, however, are not customized for smart
factories.

2.3 Protocol Design for Smart Factory

In this section, we introduce our protocol design for the smart factory use case in
IIoT [57]. We begin with introducing the considered network scenario and then
present the elements of our protocol design one by one.

2.3.1 Networking Scenario

Consider a fully connected network with one AP covering a limited geographical

area, e.g., a manufacturing facility.3 A large number of devices such as sensors,
actuators, and controllers are densely deployed in the area. The devices are
categorized into three types, i.e., high-priority (HP) devices, regular-priority (RP)
devices, and low-priority (LP) devices. An illustration of the considered scenario is
given in Fig. 2.1.
The overall number of devices and the set of devices are denoted by D and D,
respectively. The number and set of HP, RP, and LP devices are denoted by D H
and DH , D R and DR , and D L and DL , respectively. Without loss of generality, we
assume that the devices are indexed such that the first till the D H th devices are the
HP devices, the next D R devices are the RP devices, and the last D L devices are the
LP devices.
Communication Characteristics. The communication characteristics include:
• Short data packets—The length of physical-layer packets is normally in the range
between several bytes to several hundred bytes [58];
• Uplink-dominated transmission—A significant portion of the data traffic is
attributed to sensor readings or device status reports [6].
QoS Requirements The considered QoS metrics are delay, from the instant of
packet arrival to the instant of successful packet transmission, and packet trans-
mission collision probability. Different types of devices have different QoS require-
ments. Specifically, the maximum tolerable delay and packet collision probability
for HP, RP, and LP devices are denoted by δ H and ρ H , δ R and ρ R , and δ L and ρ L

3 The target area is assumed to be less than 1 km2 .

2.3 Protocol Design for Smart Factory 19

Fig. 2.1 An illustration of the networking scenario

respectively, where δ H < δ R < δ L and ρ H < ρ R < ρ L . The value of δ H is assumed
to be small such as on the millisecond level.

Device Packet Arrivals For practicality, we do not assume a specific traffic model.
However, we consider the following data packet arrival properties:
• The packet arrival statistics at each device are constant during a relatively long
period with respect to packet inter-arrival time. The packet arrival rate of device
i in the considered time duration is denoted by λi ;
• The packet arrival rate is relatively low so that 1/λi is much larger than δ H for
any i. This is in accordance with the sporadic transmission characteristic of MTC,
where the packet inter-arrival time can range from tens of milliseconds to several
minutes [7];
• For tractability, we assume that the transmission time for data packets is identical
and equal to Tx .
Given the networking scenario, we aim to develop a MAC solution with the
following features:
(1) Accommodating a large number of devices on a single channel with a single
AP;
(2) Satisfying the differentiated QoS requirements for each type of devices;
(3) Keeping control overhead as low as possible;
20 2 Industrial Internet of Things: Smart Factory

(4) Exploring the role of machine learning, specifically in device transmission

scheduling.
Our MAC protocol design is based on time-slotted channel access, which suits short
packets. Tailored for the considered networking scenario, our protocol comprises the
following elements:
• Mini-slot based carrier sensing (MsCS);
• Synchronization carrier sensing (SyncCS);
• Differentiated assignment cycles;
• Superimposed mini-slot assignment (SMsA).
The first two elements target at improving channel utilization efficiency through
implicit distributed coordination, the third targets at providing differentiated QoS
for different device types, and the last targets at increasing the number of supported
devices.

2.3.2 Mini-Slot Based Carrier Sensing (MsCS)

Time is partitioned into frames, and each frame is partitioned into ns slots, as shown
in Fig. 2.2. A slot begins with nm mini-slots, each of length Tm , followed by a
duration of length Tx . Accordingly, the length of a slot, denoted by Ts , depends
on the number of mini-slots and is equal to nm × Tm + Tx .
Given the high device density and sporadic transmission pattern, each slot is
assigned to multiple devices, in order to achieve high channel utilization efficiency
via reducing idle slots. Different devices associated with a slot are assigned different
mini-slots of the slot. Different from existing designs with mini-slots, where mini-
slots are used for transmitting packets [59, 60] or jamming signals [61], the
mini-slots in our protocol are very short, e.g., less than 10 microseconds (µs) and
are used for channel sensing instead of sending reservation requests or data packets.
In the proposed protocol, the minimum time unit for transmitting a packet is a slot,
and each slot accommodates at most one successful packet transmission. Clearly,
without proper coordination, transmission collisions may happen when multiple
devices are assigned to the same slot.
The purpose of using mini-slots is to enable channel sensing for collision-free
distributed channel access. When the AP assigns a slot to a device, it also specifies
a mini-slot for the device. Suppose that device i is assigned mini-slot m of slot l.
Then, the following rules are used in the proposed protocol:
• If device i has a packet to transmit and m = 1, it starts transmitting right away
when slot l begins;
• If device i has a packet to transmit and m > 1, it needs to sense the channel
during mini-slot m − 1 of slot l and starts transmitting from mini-slot m of slot l
only if the channel is sensed idle. Otherwise, it will skip this slot and wait for the
next transmission opportunity;
2.3 Protocol Design for Smart Factory 21

Fig. 2.2 An illustration of the frame, slot, and mini-slot structure

• If device i does not have a packet to send, it simply stays idle in the corresponding
slot.
The first two cases are illustrated in Fig. 2.3.
With MsCS, different mini-slots correspond to different transmission priorities.
Specifically, a mini-slot with a larger index corresponds to a lower transmission
priority. Therefore, mini-slots with small indexes can be used to accommodate HP
devices. Via MsCS, a device makes sure that none of the devices with higher priority
is using the channel before accessing the channel. As a result, the devices can avoid
packet collision while sharing the same slot. Note that the MsCS is fully distributed
and does not require any control message exchange, given the assignment of slots
and mini-slots to devices by the AP. The cost for avoiding collision is the overhead
of using mini-slots for sensing. Specifically, the ratio of usable packet transmission
duration over slot length is Tx /Ts .
22 2 Industrial Internet of Things: Smart Factory

Fig. 2.3 An illustration of the MsCS: (a) Devices assigned to the first mini-slot of any slot starts
transmission immediately when the slot begins, without sensing the channel; (b) devices assigned
to mini-slot m(> 1) must sense the channel during the (m − 1)th mini-slot, and starts transmission
at the beginning of the mth mini-slot if the channel is sensed to be idle

For MsCS to work, the following conditions should be satisfied:

• The mini-slot length, Tm , must be longer than the maximum propagation delay
across the network coverage area;4

4 A possible choice for mini-slot length is 9 µs, which follows from the DCF slot time in

IEEE 802.11ac.
2.3 Protocol Design for Smart Factory 23

• The overall length of all mini-slots, i.e., nm Tm , should be less than the packet
transmission duration Tx . This is for ensuring that each slot accommodates at
most one transmission;5
• The aggregated packet arrival rate of all devices assigned the same slot must be
less than 1 per frame.

2.3.3 Synchronization Sensing (SyncCS)

Even though MsCS improves channel utilization efficiency, as a result of multiple

devices sharing each slot, none of the devices may have a packet to transmit in a slot.
Increasing the number of mini-slots in each slot can reduce the slot idle probability.
However, it may violate the delay requirements for devices assigned high-index
mini-slots or the aforementioned condition that nm Tm ≤ Tx .
Alternatively, if idle slots can be identified and avoided, the channel utilization
efficiency can be further improved, and so will the resulting QoS. To achieve this,
the following rules of SyncCS are used in the proposed protocol:
• All devices in D sense the channel in the last mini-slot, i.e., mini-slot nm , of
everyone slot. The exceptions are: (i) any device that is transmitting and (ii) the
device that is assigned mini-slot nm ;6
• If the last mini-slot is idle, the rest of the current slot is skipped and the next slot
starts immediately after this last mini-slot;
• If the last mini-slot is busy, the next slot starts after the current slot ends.
The above rules are illustrated in Fig. 2.4, and the rationale is explained as follows.
Given the condition that nm Tm < Tx as mentioned in Sect. 2.3.1, no device is
or will be transmitting in a slot if the last mini-slot of that slot is idle. Therefore,
upon sensing an idle last mini-slot, all devices know that the rest of the slot can be
skipped and the next slot can start after this mini-slot. The SyncCS allows devices to
synchronize slots even though the length of a slot is no longer fixed. With SyncCS,
a busy slot has the full length of nm × Tm + Tx , while an idle slot has the reduced
length of nm × Tm .
The SyncCS has two main differences from the MsCS:
• In SyncCS, devices must perform sensing regardless of whether they have a
packet to transmit or not (with exceptions as mentioned above);
• In SyncCS, all devices, not just the devices assigned to the slot, need to sense the
channel in each slot.

5 In an extreme case when a device assigned a low-index mini-slot transmits a very short packet, it

is possible that a device assigned a high-index mini-slot senses channel idle and transmits a packet
in the same slot. This extreme case is ignored in the protocol design and performance analysis.
6 The device assigned mini-slot n knows whether the slot is idle or not from sensing the channel
m
during mini-slot nm − 1 as mandated by the MsCS.
24 2 Industrial Internet of Things: Smart Factory

Fig. 2.4 An illustration of the syncCS: (a) when the last mini-slot of a slot is sensed idle, the
remaining transmission duration of this slot is skipped, and the next slot starts right after the last
mini-slot of this slot; (b) when the last mini-slot of a slot is sensed busy, the next slot starts after
the entire duration of this slot

Similar to the MsCS, SyncCS is fully distributed and does not require any control
message exchange. The cost for further improving channel utilization efficiency via
SyncCS is the extra channel sensing. In addition, accurate time synchronization is
required among all devices. Without SyncCS, a device can be in the sleep mode
2.3 Protocol Design for Smart Factory 25

for most of the time in a frame and only wake up before its assigned mini-slot for
MsCS if it has a packet to transmit. With SyncCS, each device needs to perform
sensing in each slot and re-synchronize once for each idle slot. In the IIoT scenario
under consideration, it is possible that energy consumption of devices is less of a
concern (e.g., as compared with sensors deployed in remote areas such as in forests);
Otherwise, the design element of SyncCS can be omitted in the proposed protocol.7

2.3.4 Differentiated Assignment Cycles

Using the slot structure in Fig. 2.2, the delay for a device depends on the frame
length if each device has at most one transmission opportunity in each frame.
However, one transmission opportunity in each frame for every device does
not provide sufficient flexibility to support differentiated QoS. Particularly, the
maximum delay threshold of HP devices, i.e., δ H , can be much smaller than that
of RP/LP devices. To address this problem, we extend the frame in Fig. 2.2 to
differentiated assignment cycles. Specifically, each HP, RP, and LP assignment cycle
consists of r H , r R , and r L slots, respectively, where r H < r R < r L . Each HP, RP,
or LP device is assigned one mini-slot of one slot in each HP, RP, or LP assignment
cycle, respectively. Thus, an HP/RP/LP cycle serves as a frame for the HP/RP/LP
devices, respectively. In the case when all devices have the same priority, the HP, RP,
and LP cycles become identical and reduce to a standard frame. The differentiated
assignment cycles are illustrated in Fig. 2.5, in which different color patterns in the
mini-slots represent different assigned devices. In the illustration, r L is a multiple
of r R , and r R is a multiple of r H .8 The HP devices assigned to the same slot in any
different HP assignment cycles are identical, as shown by the two illustrated slots at
the top of Fig. 2.5, while the RP or LP devices assigned to the two slots are different.
With differentiated assignment cycles, it becomes possible to achieve the strin-
gent delay requirement of HP devices, by setting r H small, and at the same time
support a large number of devices, by using a large r R and/or r L . Note that similar
idea of differentiated cycles can be found in existing works such as [62], where two
different cycle lengths are used for realtime and non-realtime traffic, respectively.
With a different slot structure and three different cycle lengths, we adopt the same
essential idea here. This is because, for scheduling based channel access, achieving
lower delay translates to more frequently scheduled transmission opportunities. This
naturally leads to differentiated cycles for different device or traffic types.

7 Alternatively, the AP may broadcast frame synchronization beacons. In such case, when a device

has a packet to send, it can wake up and synchronize to the next frame. It may remain awake and
synchronized to each slot until the packet is transmitted.
8 While r L does not have to be a multiple of r R or r H in theory, the overall device assignment cycle

is the lowest common multiple of r H , r R , and r L . Limiting the lowest common multiple to be r R
itself can reduce the complexity of device assignment by the AP.
26 2 Industrial Internet of Things: Smart Factory

2.3.5 Superimposed Mini-slot Assignment (SMsA)

The proposed MAC protocol aims to support a high device density. The MsCS
and SyncCS contribute to the solution by improving channel utilization efficiency,
along with differentiated assignment cycles with a large r R and/or r L . In addition, if
devices can share a mini-slot, beyond only sharing a slot, the capacity of the network
in terms of the number of supported devices can be significantly improved, at the
cost of nonzero packet transmission collision probabilities.
The final element in our proposed protocol, i.e., SMsA, allows the assignment of
one mini-slot to multiple devices, provided that packet transmissions associated with
such assignment can be properly scheduled as not to violate the QoS requirements of
the devices. For the simplicity of presentation, we limit the SMsA to devices of the
same type, i.e., an HP device can share a mini-slot only with other HP devices. With
SMsA, a mini-slot in Fig. 2.5 may no longer be assigned to a device exclusively.
Transmission collision may happen among devices sharing a mini-slot, and the
collision probability depends on the following factors:
• The device packet arrival rates;
• The number of mini-slots and the mini-slot assignment;

Fig. 2.5 An illustration of differentiated assignment cycles

2.3 Protocol Design for Smart Factory 27

• The HP, RP, and LP assignment cycle lengths.

While the device packet arrival rates are not controllable, the collision probability
may be reduced by properly determining the last two factors (to be studied in
Sect. 2.5).
We do not consider collision resolution here. However, a design element for
collision detection can be added to our proposed MAC protocol. The following is
an example. If two or more devices assigned the same mini-slot simultaneously
start sending packets to the AP, the AP will detect the collision. As soon as the AP
detects the collision, it will start broadcasting a collision beacon that fills the rest of
the current slot. On the device side, the sending devices will switch to sensing mode
to check for a collision beacon after transmitting their packets. If a beacon is sensed,
the device knows that a packet collision happened during its transmission and may
decide to re-transmit the packet in another slot.

2.3.6 Downlink Control

The AP broadcasts the mini-slot and slot assignment to devices via downlink control
messages. Based on the assumption of stationary traffic statistics in a relative long
duration,9 the assignment does not need to be updated frequently. The AP may either
broadcast the entire assignment in one downlink control message or breakdown the
assignment information into multiple messages.
Consider an example of 10 mini-slots per slot (i.e., nm = 10) and 200 slots per
LP assignment cycle (i.e., r L = 200). In such case, 2 bytes is more than sufficient
to represent the slot and mini-slot assignment for each device. For 1000 devices,
the assignment message payload size is no more than 2 kilobytes (kB). For a slot
length of 200 µs, an LP assignment cycle is about 40 ms in length. Even if the
traffic statistics change as frequently as once in every 5 minutes, the 2 kB downlink
assignment message is needed just once in every 7500 LP assignment cycles or,
equivalently, 1.5 × 106 slots.
As downlink control messages are infrequent in comparison with the dominating
uplink messages, we neglect the impact of downlink control messages while
analyzing the performance of the proposed protocol.

The core of MAC protocol design is to coordinate transmissions from

devices, while prioritizing and device grouping are two important aspects

(continued)

9 The stationary duration, if denoted by Tst , should satisfy Tst  r L Ts .

28 2 Industrial Internet of Things: Smart Factory

of coordination. There are various approaches for prioritizing, such as using

both contention-based and contention-free access in a MAC protocol [63].
Similarly, there are many grouping approaches, such as limiting contention
to devices generating packets around the same instant [64]. In our proposed
MAC protocol, the utilization of mini-slots is inherently capable of both
prioritizing and grouping. Meanwhile, the differentiated assignment cycles
further strengthen the design’s capability in prioritizing, while the SMsA
further strengthens its capability in grouping.

2.4 Performance Analysis

In this section, we present performance analysis of the proposed MAC protocol,

focusing on the MsCS, SyncCS, and SMsA. Note that the proposed MAC protocol
works under the following conditions:
• The expected number of packet arrivals summarized over all devices sharing a
slot is less than 1 per frame;10
• The average packet arrival interval of any device is larger than the maximum
tolerable packet delay of that device.
In practice, some devices can have a high packet arrival rate that violates the above
conditions. In such case, more than one slot can be assigned to such a device in the
corresponding assignment cycle so that the expected number of packet arrivals of
the device per scheduled slot is less than one. In the subsequent analysis, we simply
assume that the number of packet arrivals for any device is less than one per its
assignment cycle.
Without assuming a specific traffic model, we focus on the first-order statistic.
The expected number of packet arrivals at device i in a frame is given by λi Tf ,
where Tf denotes the length of a frame. Denote the set of all HP, RP, and LP
devices assigned to slot l by Dl . Denote the delay of device i, averaged over
packet transmissions while the traffic is stationary, by τi . The aforementioned two
conditions correspond to the following equations:

λi Tf ≤ 1, ∀l, (2.1a)
i∈Dl

1
≥ τi , ∀i. (2.1b)
λi

10 This condition applies to the case without differentiated assignment cycles. In the case with

differentiated assignment cycles, the condition is different.

2.4 Performance Analysis 29

2.4.1 Delay Performance with No Buffer

We investigate the impact of mini-slots in the case without SMsA, given the slot
assignment and device packet transmission probabilities (estimated from the packet
arrival rates). Starting from a simplified scenario, the analysis here is based on the
following assumptions:
• The condition in (2.1a) is satisfied;
• A packet not in transmission is dropped when a new packet is generated. The
scenario where devices have buffers is analyzed in Sect. 2.4.2;
• All devices are of the same type and priority. Consequently, the three assignment
cycles reduce to a unified frame with ns slots;
• The SyncCS is not adopted. The analysis of SyncCS is given in Sect. 2.4.4.
We focus on the delay analysis since collision probability is zero without SMsA.
Let τ0 denote the base delay, defined as the time duration from the packet arrival
instant till the first assigned mini-slot. Under the aforementioned assumptions, the
average base delay is equal to ns Ts /2 for all devices, as each device has one assigned
mini-slot in each frame. The overall delay is the base delay plus the access delay
(AD), i.e., the duration from the first assigned mini-slot since the packet arrival till
the end of the packet transmission. Since the average base delay is a constant here,
we focus on finding the average AD.
Denote the device assigned the mth mini-slot of the lth slot by dm,l . Denote by
τm,l the average access delay counted in frames (AD-F), i.e., the number of logical
frames since device dm,l ’s packet arrival till device dm,l ’s packet transmission.11
Different from a physical frame, a logical frame counted in the AD-F for device
dm,l is the duration from slot l of one physical frame to slot l of the next physical
frame. Therefore, a logical frame has the same length as a physical frame, but
different starting and ending points for different devices. Accordingly, the arrival
and transmission of a packet can happen within one logical frame, and the resulting
AD-F is 1 in such case.12 Note that AD-F τm,l corresponds to a duration slightly
longer than the AD defined in the preceding paragraph. This is because the AD
ends when a packet transmission is completed, while the AD-F counts the entire
frame into the delay, including the duration after device dm,l ’s packet transmission.
Accordingly, the AD of device dm,l can be obtained from the AD-F by calculating
(τm,l − 1) × Tf + Tx , where the frame length Tf is equal to ns Ts .

11 When packet re-transmission is considered, the definition of AD-F should be changed to “the

number of logical frames since packet arrival till successful packet transmission”. Meanwhile, the
“packet arrival rate” in our analysis should be replaced by “packet transmission rate” as a packet
may need re-transmission(s).
12 In the rest of this chapter, we do not distinguish physical and logical frames and refer to both as

“frame” since they are equal in length.

30 2 Industrial Internet of Things: Smart Factory

Since any device assigned the first mini-slot of any slot can transmit right away
without sensing when the slot begins, we have

τ1,l = 1, ∀l. (2.2)

For devices assigned the subsequent mini-slots, the AD-F can be found using the
following result.
Theorem 2.1 For any integer m such that 1 ≤ m ≤ nm − 1, the following relation
between the AD-F of device dm+1,l and device dm,l holds:

1 (1 − γm,l )Tf λm,l
τm+1,l = − 2
τm,l
1 − γm,l − Tf λm,l 2

  Tf λm,l (1+γm,l )
+ 1−γm,l +Tf λm,l τm,l − (2.3)
2

where
λm,l
λm,l =   (2.4)
1 + Tf λm,l (τm,l − 1/2)

represents the effective packet arrival rate of device dm,l excluding dropped packets
due to packet replacement (as there is no buffer), and


m
γm,l = Tf λr,l (2.5)
r=1

represents the expected overall number of packet arrivals in a frame for devices d1,l
till dm,l (excluding replaced packets).
The proof of the above theorem can be found in [57]. Using the fact that τ1,l is
equal to 1 for any l, (2.3) can be used to obtain the AD-F for devices assigned to all
subsequent mini-slots in a slot recursively.

2.4.2 Delay Performance with Buffer

Now consider the case when each device has a buffer. Recall that different mini-
slots correspond to different transmission priorities. In the proposed protocol, any
proper slot and mini-slot assignment ensures that the expected number of packets in
the buffer of device dm,l is less than one, for any m < nm and any l. The reason is
that, if the expected number of buffered packet at dm,l is larger than or equal to one,
devices assigned mini-slots m + 1, . . . , nm of slot l have almost no opportunity to
2.4 Performance Analysis 31

transmit. As a result, we neglect the case when there are more than one packet in a
buffer and use the following approximation. Specifically, at any instant, a device is
in one of three states:
• no packet;
• one packet, transmitting or waiting for channel access;
• two packets, one transmitting or waiting for channel access and the other arriving
and going into the buffer.
Accordingly, for any given device, there is either no packet or one packet transmit-
ting or waiting for channel access when a new packet arrives.
Denote by τm,lb the average AD-F of device d
m,l in the case with buffer, the
following result is in order.
Theorem 2.2 In the case with buffers, for any integer m such that 1 ≤ m ≤ nm − 1,
the relation between the AD-F of device dm+1,l and device dm,l is given by

b
  b )T λ
1−γm,l 1 (1−γm,l f m,l
b
τm+1,l = b b −T λ
−
1−γm+1,l 1−γm,l f m,l 2

 2 b )
Tf λm,l (1+γm,l
· b
τm,l +(1−γm,l +Tf λm,l )τm,l −
b b
− 1 +1
2

where


m
b
γm,l = Tf λr,l (2.6)
r=1

represents the expected overall number of packet arrivals in a frame for devices d1,l
till dm,l .
The proof of the above theorem can be found in [57].

2.4.3 Slot Idle Probability

A slot is idle if none of its associated devices transmits. Under stationary packet
arrival statistics, the expected slot idle probability of MsCS can be obtained. In the
case with and without buffer, the slot idle probability is approximately given by


nm
ηlb = 1 − λm,l Tf (2.7a)
m=1
32 2 Industrial Internet of Things: Smart Factory


nm
ηl = 1 − λm,l Tf (2.7b)
m=1

respectively, where λm,l is given in (2.4). Note that the right-hand side of either of
the two equations above is non-negative when the condition (2.1a) is satisfied, i.e.,
when the slot is not overloaded. The above approximation of slot idle probability
also assumes a negligible packet collision probability, i.e., the expected number of
transmitted packets and the expected number of packet arrivals (that cause no packet
replacement) are equal in any slot.
Define the throughput of slot l as the expected number of packets transmitted in
the slot. The slot throughput equals 1 − ηlb and 1 − ηl for the cases with and without
buffers, respectively.

2.4.4 Impact of SyncCS

As SyncCS results in two possible lengths for each slot, i.e., the full and the reduced
lengths, the frame length becomes a random variable. Denote the expected frame
length with SyncCS in the case with and without buffer by Tfe,b and Tfe , respectively.
Denote ns as the number of busy slots out of the ns slots in a frame. In the case
without buffer, it follows that

Tfe = ns nm Tm + ns Tx . (2.8)

Since there is no collision,


Tfe λm,l = ns (2.9)
l m

because the expected number of packet transmissions should equal the expected
number of arriving packets (that are not replaced) in a frame duration. From (2.8),
(2.9), and (2.4) (with Tf replaced by Tfe ), ns and Tfe can be solved.
In the case with buffer, we have

Tfe,b = ns nm Tm + ns Tx (2.10a)


Tfe,b λm,l = ns (2.10b)
l m

which gives

ns nm Tm
Tfe,b = . (2.11)
1− λm,l Tx
l m
2.4 Performance Analysis 33

Substituting Tf in (2.3) and (2.6) with Tfe and Tfe,b , respectively, gives the AD-
F of the proposed design with MsCS and SyncCS. In the case without buffer, Tfe
depends on τm,l through (2.4), which renders a complicated relation.

2.4.5 Impact of SMsA

The AD-F in Sects. 2.4.1 and 2.4.2 is obtained when each mini-slot is assigned to a
device exclusively. With SMsA, we have the following questions:
• What is the relation among the AD-F of different devices assigned the same
mini-slot?
• How does the SMsA impact the relation in the AD-F between devices assigned
adjacent mini-slots?
Denote the set of all devices assigned mini-slot m of slot l by Dm,l . The following
theorem answers the first question.
Theorem 2.3 In the case without buffer, all devices in Dm,l have the same AD-F,
regardless of the difference in their individual packet arrival rates. In the case with
buffer, assuming a negligible packet collision probability and


m 
λi λj , ∀i ∈ Dm,l , (2.12)
r=1 j ∈Dr,l

the differences among the AD-Fs of devices in Dm,l are negligible.

For the second question, similar to (2.3) and (2.6), the relation between the AD-
Fs of devices in adjacent mini-slots in the case of SMsA can be characterized. The
characterization is given in [57]. It is worth mentioning that the packet collision
probability has an impact on the AD-F even if devices do not detect collisions or
re-transmit. Given the aggregated packet arrival rate of devices sharing a mini-slot,
a higher collision probability implies less channel busy duration for transmitting
the same amount of packets. Consequently, the average packet waiting time and
the AD-F decrease as the collision probability increases. However, if the collision
probability is low, such impact can be negligible.
With the AD-F, we can estimate the packet collision probability. Consider the
case with buffer as an example and assume that the condition in (2.12) is satisfied.
s,b
Based on Theorem 2.3, all the devices in Dm,l have the same AD-F, denoted by τm,l .
Then, any device in Dm,l with a packet to send is expected to have one transmission
s,b
opportunity in every τm,l frames. The expected number of packet arrivals at device
i ∈ Dm,l between any two consecutive transmission opportunities, which must
s,b
be less than 1, can be estimated by τm,l Tf λi . With the MsCS, all devices in Dm,l
that have packets to send share the same transmission opportunities. Therefore, the
Another Random Scribd Document
with Unrelated Content
Rosetter Gleason Cole (1866)
 Some Opera Writers

Italian

Claudio Merulo (1533–1604) (Madrigal Play)

Vincenzo Galilei (1533–1591)
Giulio Caccini (1546–1618)
Emilio del Cavalieri (1550–1602)
Jacopo Perti (1561–1633)
Claudio Monteverde (1567–1643)
Francesco Manelli (1595–1670)
Benedetto Ferrari (1597–1681)
Francesco Cavalli (1602–1676)
Giacomo Carissimi (1604–1674)
Marc Antonio Cesti (1620–1669)
Giovanni Legrenzi (1625–1690)
Domenico Gabrieli (1640–1690)
Alessandro Stradella (1645–1681)
Alessandro Scarlatti (1659–1725)
Jacopo Antonio Perti (1661–1756)
Attilio Ariosti (1666–1740)
Antonio Lotti (1667–1740)
Giuseppe Porsile (1672–1750)
Marc Antonio Buononcini (1675?–1726)
Antonio Caldara (1678?–1736)
Niccolo Antonio Porpora (1686–1766)
Leonardo Vinci (1690–1730)
Leonardo Leo (1694–1744)
Giovanni Battista Pergolesi (1710–1736)
Nicola Jommelli (1714–1774)
Nicolo Piccinni (1728–1800)
Giuseppe Sarti (1729–1802)
Antonio Sacchini (1734–1786)
Giovanni Paisiello (1741–1816)
Domenico Cimarosa (1749–1801)
Antonio Salieri (1750–1825)
Niccolo Vaccai (1790–1848)
Gioachino Antonio Rossini (1792–1868)
G. Saverio Mercadante (1795–1870)
Giovanni Pacini (1796–1867)
Gaetano Donizetti (1797–1848)
Vincenzo Bellini (1801–1835)
Errico Petrella (1813–1877)
Giuseppe Verdi (1813–1901)
Carlo Perdrotti (1817–1893)
Alberto Randegger (1833–1911)
Amilcare Ponchielli (1834–1886)
Filippo Marchetti (1835–1902)
Arrigo Boito (1842–1918)
Luigi Mancinelli (1848–1921)
Niccolo Ravera (1851)
Alfredo Catalani (1854–1893)
Giacomo Puccini (1858–1924)
Ruggiero Leoncavallo (1858–1919)
Sylvio Lazzari (1858) (Living in Paris)
Alberto Franchetti (1860)
Pietro Mascagni (1863)
Franco Leoni (1864)
Crescenzo Buongiorno (1864–1903)
Giacomo Orefice (1865–1923)
Ferruccio Busoni (1866–1924)
Francesco Cilèa (1866)
Umberto Giordano (1867)
Giovanni Gianetti (1869)
Antonio Luzzi (1873)
Italo Montemezzi (1875)
Domenico Monleone (1875)
Ermanno Wolf-Ferrari (1876)
Franco Alfano (1877)
Stefano Donaudy (1879)
Ottorino Respighi (1879)
Alberto Iginio Randegger (Austrian) (1880–1918)
Ildebrando Pizzetti (1880)
G. Francesco Malipiero (1882)
Giuseppe Gino Marinuzzi (1882)
Riccardo Zandonai (1883)
Francesco Santoliquido (1883)
Adriano Lualdi (1887)
Vincenzo Davico (1889)
Ettore Lucatello (?)
Victor de Sabata (1892)

Spanish

Salvador Giner (1832–1911)

Felipe Pedrell (1841–1922)
Tomas Breton y Hernandez (1850)
Amedeo Vives (?)
Costa Nogueras (?)
Isaac Albeniz (1861–1909)
Enrique Granados (1867–1916)
Manuel de Falla (1876)
Joan Manen (1883)
Maria Rodrigo (1888)

South America

Carlos Gomez (Brazilian) (1839–1896)

José Valle-Riestra (Peruvian) (1859)
Antonio Berutti (Argentine) (1862)

French

Robert Cambert (1628–1679) (Wrote with Perrin)

Jean Batiste Lully (Italian) (1632–1687)
Pascal Colasse (1649–1709)
Marin Marais (1656–1728)
André Campra (1660–1744)
André Destouches (1672–1749)
Jean Philippe Rameau (1683–1764)
Egidio Romualdo Duni (Italian) (1709–1775)
Jean Jacques Rousseau (1712–1778)
François André Danican Philidor (1726–1795)
Pierre Berton (1727–1780)
Pierre Alexandre Monsigny (1729–1817)
François Joseph Gossec (1743–1829)
André Ernest Modeste Grétry (1741–1813)
Luigi Cherubini (Italian) (1760–1842)
Jean François Lesueur (1760–1837)
Étienne Henri Méhul (1763–1822)
Gasparo Luigi Pacifico Spontini (Italian) (1774–1851)
François Adrien Boieldieu (1775–1834)
Nicolo Isouard (1775–1818)
Daniel François Esprit Auber (1782–1871)
Louis Joseph Ferdinand Hérold (1791–1833)
Jacques F. F. E. Halevy (1799–1862)
Adolphe Adam (1803–1856)
Hector Berlioz (1803–1869)
Napoleon Henri Reber (1807–1880)
Félicien C. David (1810–1876)
Ambroise Thomas (1811–1896)
Louis Lacombe (1818–1884)
Charles François Gounod (1818–1893)
Jacques Offenbach (German) (1819–1880)
Victor Massé (1822–1884)
Edouard Lalo (1823–1892)
Louis Ernest Reyer (1823–1909)
Florimond Ronger Hervé (1825–1892)
Alexandre Charles Lecocq (1832–1918)
Camille Saint-Saëns (1835–1921)
Ernest Guiraud (American) (1837–1892) (Born in New Orleans)
Théodore Dubois (1837–1924)
Georges Bizet (1838–1875)
Clément P. Léo Délibes (1839–1891)
Félix Victorin de Joncières (1839–1903)
Emmanuel Chabrier (1841–1894)
Jules Massenet (1842–1912)
Emile Paladilhe (1844)
Joseph Arthur Coquard (1846–1910)
Robert Planquette (1848–1903)
Benjamin Godard (1849–1895)
Augusta Holmès (British) (1849–1903)
François Thomé (1850–1905)
Alexandre Georges (1850)
Vincent d’Indy (1851)
Hillemacher Frères (Brothers) Paul (1852) Lucien (1860–1909)
Raoul Pugno (1852–1914)
Andre Messager (1853)
Samuel A. Rousseau (1853–1904)
Ernest Chausson (1855–1899)
Alfred Bruneau (1857)
Georges Hüe (1858)
Gustave Charpentier (1860)
Gabriella Ferrari (1860)
Pierre de Bréville (1861)
Claude Achille Debussy (1862–1918)
Xavier Leroux (1863–1919)
Camille Erlanger (1863–1919)
Gabriel Pierné (1863)
Paul Vidal (1863)
Alfred Bachelet (1864)
Alberic Magnard (1865–1914)
Paul Dukas (1865)
Albert Roussel (1869)
Florent Schmitt (1870)
Henri Büsser (1872)
Deodat de Sévérac (1873–1921)
Henri Rabaud (1873)
Reynaldo Hahn (Venezuelan) (1874)
Max d’Ollone (1875)
Antoine Mariotte (1875)
Maurice Ravel (1875)
Raoul Laparra (1876)
Jean Nouguès (1876)
Henri Février (1876)
Louis Aubert (1877)
Gabriel Dupont (1878–1914)
Félix Fourdrain (1880–1924)
Albert Wolf
Roland Manuel (1891)
Arthur Honegger (1892)
Darius Milhaud (1892)

Belgian

André Ernest Modeste Grétry (1741–1813) (Lived in Paris)

François Auguste Gevaert (1828–1908)
Pierre Benoit (1834–1901)
Jan Blockx (1851–1912)
Edgar Tinel (1854–1912)
Valentin Neuville (1863)
Paul Gilson (1865)
Alfred Kaiser (1872)
Charles Radoux (1877)
Albert Dupuis (1877)
Joseph van der Meulen (?)

Dutch

Anton Berlijn (1817–1870)

Johan Wagenaar (1862)
Jan Brandt-Buys (1868)
Cornelis Dopper (1870)
Charles Grelinger (?)

Swiss

Emile Jacques-Dalcroze (1865)

Gustave Doret (1866)
Othmar Schoeck (1886)

German

Heinrich Schütz (1585–1672)

Johann Thiele (1646)
Johann Christoph Pepusch (1667–1752) (Lived in London)
Johann Conradi (1670?–?)
Reinhard Keiser (1674–1739)
Johann Mattheson (1681–1764)
George Frederick Handel (1685–1759)
Johann Adolph Hasse (1699–1783)
Karl Heinrich Graun (1701–1759)
Christoph Willibald Gluck (1714–1787)
Johann Adam Hiller (1728–1804)
Karl Ditters von Dittersdorf (1739–1799)
Abt (Georg Joseph) Vogler (1749–1814)
Johann Friedrich Reichardt (1752–1814)
Peter von Winter (1754–1825)
Wolfgang Amadeus Mozart (1756–1791)
Joseph Weigl (1766–1846)
Ludwig van Beethoven (1770–1827)
Christoph Ernst Friederich Weyse (Danish) (1774–1842)
Konradin Kreutzer (1780–1849)
Ludwig (Louis) Spohr (1784–1859)
Carl Maria von Weber (1786–1826)
Peter Joseph von Lindpainter (1791–1856)
Giacomo Meyerbeer (1791–1864) (Lived in Paris)
Heinrich Marschner (1795–1861)
Karl Gottlieb Reissiger (1798–1859)
Gustav Albert Lortzing (1801–1851)
Otto Nicolai (1810–1849)
Friedrich von Flotow (1812–1883)
Richard Wagner (1813–1883)
Franz von Suppé (Austrian) (1819–1895)
Peter Cornelius (1824–1874)
Felix Draeseke (1835–1913)
Max Zenger (1837–1911)
Hermann Goetz (1840–1876)
Victor Nessler (Alsatian) (1841–1890)
Heinrich Hofmann (1842–1902)
Ignaz Brüll (Austrian) (1846–1907)
August Bungert (1846–1915)
August Klughardt (1847–1902)
Cyril Kistler (1848–1907)
Ivan Knorr (1853–1916)
Engelbert Humperdinck (1854–1921)
Heinrich Zöllner (1854)
Paul Geisler (1856–1919)
Wilhelm Kienzl (1857)
Alexander von Fielitz (1860)
Maria Joseph Erb (Alsatian) (1860)
Emil Nikolaus Reznicek (Austrian) (1861)
Ludwig Thuille (1861–1907)
Hugo Kaun (1863)
Felix Weingartner (1863)
Richard Strauss (1864)
Waldemar von Bausznern (1866)
Max Schillings (1868)
Hans Pfitzner (1869)
Siegfried Wagner (1869)
Alexander von Zemlinsky (Austrian) (1872)
Leo Fall (Austrian) (1873–1925)
Paul Graner (1873)
Waldemar Wendland (1873)
Julius Bittner (1874)
Arnold Schoenberg (1874)
Joseph Gustav Mrazek (Austrian) (1878)
Franz Schreker (Austrian) (1878)
Edgar Istel (1880)
Walter Braunfels (1882)
Egon Wellesz (Austrian) (1885)
Alban Berg (Austrian) (1885)
Paul Hindemith (1895)
Erich Korngold (Austrian) (1897)
Karl Krafft-Lortzing (?–1923)
Ignatz Waghalter (20th Century)

Czecho-Slovakian

Bedrich (Friedrich) Smetana (1824–1884)

Eduard Napravnik (1839–1916)
Antonin Dvorak (1841–1904)
Zdenko Fibich (1850–1900)
Leos Janacek (1855–1928)
Hans Trnecek (1858–1914)
Josef Bohuslav Foerster (1859)
Emil Nikolaus Reznicek (1861)
Karl Kovarovic (1862–1920)
Karel Weis (1862)
Stanislaus Suda (1865)
Karl Navratil (1867)
Vitezslav Novak (1870)
Adolf Piskacek (1874)
Camillo Hildebrand (1876)
Otakar Ostrcil (1879)
Otakar Zich (1879)
Rudolf Karel (1881)
Bohuslav Martinu (1890)

Hungarian

Karl Goldmark (1830–1915)

Alphons Czibulka (1842–1894)
Jeno Hubay (1858)
Julius J. Major (1859)
Emanuel Moor (1862)
Georg Jarno (1868–1920)
Béla Bártok (1881)
Emil Abrányi (1882)

Scandinavian

Johan P. E. Hartmann (Danish) (1805–1900)

Ivar Hallstrom (Swedish) (1826–1901)
Eduard Lassen (Danish) (1830–1904)
Emil Hartmann (Danish) (1836–1898)
Anders Hallen (1846)
Peter Lange-Müller (Danish) (1850–1926)
Christian Sinding (Norwegian) (1856)
Gerhard Schjelderup (Norwegian) (1859)
August Enna (Danish) (1860)
Wilhelm Stenhammar (Swedish) (1871)
Hakon Boerresen (Danish) (1876)
Paul August von Klenau (Danish) (1883)
Ture Rangstrom (Swedish) (1884)

Finnish

Jan Sibelius (1865)

Selim Palmgren (1878)
Armas E. Launis (1884)

Russian

Michail I. Glinka (1804–1857)

Alexander S. Dargomyzsky (1813–1869)
Alexander Serov (1820–1871)
Anton Rubinstein (1830–1894)
Alexander P. Borodin (1834–1887)
César Cui (1835–1918)
Modest Moussorgsky (1839–1881)
Peter I. Tchaikovsky (1840–1893)
Nikolai Rimsky-Korsakov (1844–1908)
Michail Ippolitov-Ivanov (1859)
Anton Arensky (1861–1906)
Alexander Gretchaninov (1864)
Alexander Glazounov (1865)
Vladimir Rebikov (1866)
Arseni Korestchenko (1870)
Serge Rachmaninov (1873)
Igor Stravinsky (1882)
Serge Prokofiev (1891)

Polish

Ladislas Zelenski (1837–1921)

Sigismund Noskowski (1846–1909)
Ignace Jan Paderewski (1860)
Karol Szymanowski (1883)
Ludomir von Rozycki (1883)
Raoul Koczalski (1885)

Roumanian

Theodor Flondor (? 1908)

English

Thomas Campion (1575–1620)

John Coperario (1580)
William Lawes (1582–1645)
Henry Lawes (1596–1662)
John Banister (1630–1679)
Matthew Lock (1632?–1677)
Henry Purcell (1658–1695)
Thomas Arne (1710–1778)
William Shield (1748–1829)
Stephen Storace (1763–1796)
Henry R. Bishop (1786–1855)
John Barnett (1802–1890)
Julius Benedict (1804–1885)
Michael William Balfe (1808–1888)
George A. MacFarren (1813–1887)
William Vincent Wallace (1814–1865)
Frederic Clay (French) (1838–1889)
Arthur Sullivan (1842–1900)
Alfred Cellier (1844–1891)
Alexander Campbell Mackenzie (1847)
A. Goring Thomas (1851–1892)
Charles Villiers Stanford (1852–1924)
Frederick Corder (1852)
Frederic Hymen Cowen (Jamaica) (1852)
Ethel Mary Smyth (1858)
Isidore De Lara (1858)
Marie Wurm (1860) (Living in Germany)
Liza Lehmann (1862–1918)
Edward German (1862)
Frederick Delius (1863)
Eugene d’Albert (1863) (Living in Germany)
Edward Woodall Naylor (1867)
Gustav Holst (1874)
Josef Holbrooke (1878)
Albert Coates (Russian) (1882)
Hubert Bath (1883)
Lord Berners (1883)

American

William H. Fry (1813–1864)

George Bristow (1825–1898)
John Knowles Paine (1839–1906)
Frederic Grant Gleason (1848–1903)
William J. McCoy (1848)
Max Vogrich (Transylvanian) (1852–1916)
George W. Chadwick (1854)
Julian Edwards (English) (1855–1910)
Humphrey John Stewart (English) (1856)
Richard Henry Warren (1859)
Reginald De Koven (1859–1920)
Victor Herbert (Irish) (1859–1924)
Pietro Floridia (Italian) (1860)
Walter Damrosch (1862)
Horatio W. Parker (1863–1919)
Ernest Carter (1866)
N. Clifford Page (1866)
Paolo Gallico (Austrian) (1868)
Wallace A. Sabin (English) (1869)
Louis Adolphe Coerne (1870–1922)
Joseph C. Breil (1870–1926)
Henry K. Hadley (1871)
Frederick Converse (1871)
Arthur Nevin (1871)
Mary Carr Moore (1873)
Theodore Stearns (?)
Frank Patterson (?)
John Adam Hugo (1873)
Albert Mildenberg (1878–1918)
Ernest Bloch (Swiss) (1880)
Charles Wakefield Cadman (1881)
Simon Buchhalter (Russian) (1881)
Lazare Saminsky (Russian) (1883)
Louis Gruenberg (1884)
W. Frank Harling (?)
Isaac Van Grove (?)
Timothy Spelman (1891)
Eugene Bonner (?)
 Modern Ballets

Russian

Peter I. Tchaikovsky (1840–1893)

Nikolai Rimsky-Korsakov (1844–1908)
Alexander Glazounov (1865)
Arseni Korestchenko (1870)
Nikolai Tcherepnin (1873)
Igor Stravinsky (1882)
Maximilian Steinberg (1883)
Serge Prokofiev (1891)

Polish

Karol Szymanowsky (1883)

Alexandre Tansman (20th Century)

French

Léo Délibes (1836–1891)

Emanuel Chabrier (1841–1894)
Jules Massenet (1842–1912)
Vincent d’Indy (1851)
André Messager (1853)
Alfred Bruneau (1857)
Gabriel Pierné (1863)
Erik Satie (1866–1925)
Charles Silver (1868)
Florent Schmitt (1870)
Jean Roger Ducasse (1875)
Maurice Ravel (1875)
Louis Aubert (1877)
Louis Durey (1888)
Roland Manuel (1891)
Darius Milhaud (1892)
Arthur Honegger (1892)
Germaine Taillefer (1892)
George Auric (1898)
François Poulenc (1898)
Henri Sauguet (20th Century)

Italian

Franco Alfano (1877)

Ottorino Respighi (1879)
Ildebrando Pizzetti (1880)
G. Francesco Malipiero (1882)
Alfredo Casella (1883)
Vittoria Rieti (1898)

Spanish

Manuel de Falla (1876)

Hungarian

Béla Bártok (1881)

Scandinavian

Kurt Atterberg (Swedish) (1887)

Austrian

Arnold Schoenberg (1874)

Egon Wellesz (1885)

Czecho-Slovakian
Karl Kovarovic (1862–1920)
Georg Kosa (1897)
Bohuslav Martinu (1890)

English

Arthur Seymour Sullivan (1842–1900)

Ralph Vaughn Williams (1872)
Gustav Holst (1874)

American

Henry F. Gilbert (1876–1928)

John Alden Carpenter (1876)
Charles Griffes (1884–1920)
Emerson Whithorne (1884)
Deems Taylor (1885)
Cole Porter
Leo Sowerby (1895)
 INDEX

A
Absolute music, 239–40, 242, 397, 422
Abt, Franz, 423–4
Adam de la Halle, 102, 112, 125
Aida, Verdi’s, 379–81
Albéniz, Isaac, 453–4
Alcuin, 76
Alfred the Great, 93
Alphonso XII of Spain, 453
Ambrose, St., 71, 72
America, see United States
American Academy in Rome, 507–8
American composers, 475 ff.
American folk music, 140–5
American Music Guild, the, 506–7
American opera companies, 514
American patrons of music, 512–13
American song writers, recent, 509–10
American symphony orchestras, 513–14
Anglican church, founding of, 188
Anglin, Margaret, 469
Antiphony, use of, by the Greeks, 41;
introduction into church music, 70
Apollo, 33–4
Arabia, music of, 55 ff., 209, 210;
the Arab scales, 58–9;
instruments of, 59–61
Arcadelt, Jacob, 157
Armide, Dvorak’s, 447
Arne, Dr. Thomas, 200, 339
Assyrian music, 24–5
Atonality, 517, 529
Auber, Daniel François Esprit, 333–4
Aulos of the Greeks, 42–3
Austrian National Hymn, written by Haydn, 282
Automatic pianos, 316–19
Aztecs, music of the, 53–4
B
Bach, Johann Christian, 254
Bach, Johann Christoph, 254
Bach, Johann Sebastian, 208, 211, 238, 240;
account of his life, 244–50;
his works, 250–3;
his sons, 253–4;
comparison with Handel, 255–6
Bach, Karl Philip Emanuel, 249, 253–4
Bach, Wilhelm Friedemann, 253
Bach Festival, yearly, at Bethlehem, Pa., 252, 464
Bagpipes, the Roman tibia, 45;
use of, by the Hindus, 66;
of the Bohemians, 135;
 of Scotland, 138
Baif, Jean Antoine, his club of poets and musicians in France, 177
Balakirev, Mily, 444–5
Balfe, Michael William, 341
Ballad, the, and the ballet, 122
Ballet, the, at the French court in second half of the 16th century, 178
Band, the difference between, and an orchestra, 234
Bantock, Granville, 543
Barber of Seville, Rossini’s, 337
Bards of ancient Britain, 89–91
Barnby, Joseph, 340
Bartlett, Homer W., 490
Bartok, Béla, 536–7
Bauer, Marion, 507
Bax, Arnold, 544
Bay Psalm Book, the, 458
Bayreuth, 371–2, 373
Beach, Mrs. H. H. A., 480–1
Beaumont and Fletcher, 173
Bede, the venerable, 75–6, 92
Beethoven, Ludwig van, 293 ff.;
account of his life, 295–302;
his friendships, 298–9;
The Moonlight Sonata, 300, 304;
his three periods, and works during, 301–2;
his opera Fidelio, 302, 305, 306, 326;
influence upon the growth of music, 303–5;
as a composer of instrumental music, 305–6;
his preference in pianos, 313;
the Kreutzer Sonata, 324;
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebooknice.com