Digital Twin-enabled AI Enhancement in Smart Critical Infrastructures for 5G

Keke Gai; Qiang Xiao; Meikang Qiu; Guolei Zhang; Jianyu Chen; Yihang Wei; Yue Zhang

doi:10.1145/3526195

Digital Twin-enabled AI Enhancement in Smart Critical Infrastructures for 5G

Keke Gai, Qiang Xiao, Meikang Qiu^*, Guolei Zhang, Jianyu Chen, Yihang Wei, Yue Zhang

^*Corresponding author for this work

School of Cyberspace Science and Technology

Research output: Contribution to journal › Article › peer-review

14 Citations (Scopus)

Abstract

Artificial Intelligence (AI) technology has been empowered to be a significant driven force within the edge context for powering up contemporary complex systems, such as smart critical infrastructure. Interconnectivity between physical and cyber spaces further introduces the needs of digital twin, which allows AI-based solutions to optimize various tasks in physical operations. However, due to the complexity of the setting in digital twin, task allocation is encountering multiple challenges, such as concurrent meeting the requirements of energy saving, efficiency, and accuracy. In this work, we propose a Digital Twin-Enabled Edge AI (DTE2AI), supported by our Energy-aware High Accuracy Strategy (EAHAS), which focuses on optimizing the training accuracy of AI tasks under the limits of training time and energy consumption. The average of the training accuracy was enhanced 12% based on our experiment evaluations.

Original language	English
Article number	45
Journal	ACM Transactions on Sensor Networks
Volume	18
Issue number	3
DOIs	https://doi.org/10.1145/3526195
Publication status	Published - 15 Sept 2022

Keywords

5G
Digital twin
Edge AI
smart critical infrastructure
task allocation

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1145/3526195

Cite this

@article{48ad3951de7f46e18fc35598e5eef9b1,

title = "Digital Twin-enabled AI Enhancement in Smart Critical Infrastructures for 5G",

abstract = "Artificial Intelligence (AI) technology has been empowered to be a significant driven force within the edge context for powering up contemporary complex systems, such as smart critical infrastructure. Interconnectivity between physical and cyber spaces further introduces the needs of digital twin, which allows AI-based solutions to optimize various tasks in physical operations. However, due to the complexity of the setting in digital twin, task allocation is encountering multiple challenges, such as concurrent meeting the requirements of energy saving, efficiency, and accuracy. In this work, we propose a Digital Twin-Enabled Edge AI (DTE2AI), supported by our Energy-aware High Accuracy Strategy (EAHAS), which focuses on optimizing the training accuracy of AI tasks under the limits of training time and energy consumption. The average of the training accuracy was enhanced 12% based on our experiment evaluations.",

keywords = "5G, Digital twin, Edge AI, smart critical infrastructure, task allocation",

author = "Keke Gai and Qiang Xiao and Meikang Qiu and Guolei Zhang and Jianyu Chen and Yihang Wei and Yue Zhang",

note = "Publisher Copyright: {\textcopyright} 2022 Association for Computing Machinery.",

year = "2022",

month = sep,

day = "15",

doi = "10.1145/3526195",

language = "English",

volume = "18",

journal = "ACM Transactions on Sensor Networks",

issn = "1550-4859",

publisher = "Association for Computing Machinery (ACM)",

number = "3",

}

TY - JOUR

T1 - Digital Twin-enabled AI Enhancement in Smart Critical Infrastructures for 5G

AU - Gai, Keke

AU - Xiao, Qiang

AU - Qiu, Meikang

AU - Zhang, Guolei

AU - Chen, Jianyu

AU - Wei, Yihang

AU - Zhang, Yue

PY - 2022/9/15

Y1 - 2022/9/15

N2 - Artificial Intelligence (AI) technology has been empowered to be a significant driven force within the edge context for powering up contemporary complex systems, such as smart critical infrastructure. Interconnectivity between physical and cyber spaces further introduces the needs of digital twin, which allows AI-based solutions to optimize various tasks in physical operations. However, due to the complexity of the setting in digital twin, task allocation is encountering multiple challenges, such as concurrent meeting the requirements of energy saving, efficiency, and accuracy. In this work, we propose a Digital Twin-Enabled Edge AI (DTE2AI), supported by our Energy-aware High Accuracy Strategy (EAHAS), which focuses on optimizing the training accuracy of AI tasks under the limits of training time and energy consumption. The average of the training accuracy was enhanced 12% based on our experiment evaluations.

AB - Artificial Intelligence (AI) technology has been empowered to be a significant driven force within the edge context for powering up contemporary complex systems, such as smart critical infrastructure. Interconnectivity between physical and cyber spaces further introduces the needs of digital twin, which allows AI-based solutions to optimize various tasks in physical operations. However, due to the complexity of the setting in digital twin, task allocation is encountering multiple challenges, such as concurrent meeting the requirements of energy saving, efficiency, and accuracy. In this work, we propose a Digital Twin-Enabled Edge AI (DTE2AI), supported by our Energy-aware High Accuracy Strategy (EAHAS), which focuses on optimizing the training accuracy of AI tasks under the limits of training time and energy consumption. The average of the training accuracy was enhanced 12% based on our experiment evaluations.

KW - 5G

KW - Digital twin

KW - Edge AI

KW - smart critical infrastructure

KW - task allocation

UR - http://www.scopus.com/inward/record.url?scp=85140836290&partnerID=8YFLogxK

U2 - 10.1145/3526195

DO - 10.1145/3526195

M3 - Article

AN - SCOPUS:85140836290

SN - 1550-4859

VL - 18

JO - ACM Transactions on Sensor Networks

JF - ACM Transactions on Sensor Networks

IS - 3

M1 - 45

ER -

Digital Twin-enabled AI Enhancement in Smart Critical Infrastructures for 5G

Abstract

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this