TY - GEN
T1 - Optimizing Task Distribution in Edge Computing Systems through Reinforcement Learning
T2 - 2024 IEEE Middle East Conference on Communications and Networking, MECOM 2024
AU - Zhang, Shengning
AU - Sambo, Yusuf
AU - Imran, Muhammad Ali
AU - Ahmad, Wasim
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Edge computing presents a promising solution for handling the growing demands of computationally intensive applications by offloading tasks to servers at the edge of the network. Dynamic offloading in real-world scenarios is under-researched and faces significant challenges. In this paper, we propose a task-offloading framework that optimizes the distribution of computational tasks between local devices and edge servers. The proposed framework leverages Deep Q-Learning (DQL) to learn optimal offloading strategies based on real-world data, taking into account latency and energy consumption. To demonstrate the effectiveness of the proposed approach, we apply the offloading framework to a video object tracking system that involves computationally intensive tasks like background subtraction, object detection, movement filtering, and Kalman filtering.Experimental results show significant improvements, with up to 47.11% reduction in latency and 79.67% reduction in energy consumption, demonstrating the effectiveness of our method in real-world scenarios.
AB - Edge computing presents a promising solution for handling the growing demands of computationally intensive applications by offloading tasks to servers at the edge of the network. Dynamic offloading in real-world scenarios is under-researched and faces significant challenges. In this paper, we propose a task-offloading framework that optimizes the distribution of computational tasks between local devices and edge servers. The proposed framework leverages Deep Q-Learning (DQL) to learn optimal offloading strategies based on real-world data, taking into account latency and energy consumption. To demonstrate the effectiveness of the proposed approach, we apply the offloading framework to a video object tracking system that involves computationally intensive tasks like background subtraction, object detection, movement filtering, and Kalman filtering.Experimental results show significant improvements, with up to 47.11% reduction in latency and 79.67% reduction in energy consumption, demonstrating the effectiveness of our method in real-world scenarios.
KW - Edge Computing
KW - Reinforcement Learning
KW - Task Distribution
KW - Task Offloading
UR - https://www.scopus.com/pages/publications/86000235834
U2 - 10.1109/MECOM61498.2024.10881341
DO - 10.1109/MECOM61498.2024.10881341
M3 - Conference contribution
AN - SCOPUS:86000235834
T3 - 2024 IEEE Middle East Conference on Communications and Networking, MECOM 2024
SP - 452
EP - 457
BT - 2024 IEEE Middle East Conference on Communications and Networking, MECOM 2024
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 17 November 2024 through 20 November 2024
ER -