Towards Energy-Efficiency: Integrating MATD3 Reinforcement Learning Method for Computational Offloading in RIS-Aided UAV-MEC Environments

作     者：Wu, Liangshun Zhang, Cong Zhang, Bin Du, Jianbo Qu, Junsuo 

作者机构：Key Laboratory of Embedded System and Service Computing Tongji University Ministry of Education Shanghai201804 China Shanghai Key Laboratory of Trustworthy Computing East China Normal University Shanghai201210 China School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University Shanghai200240 China Xi'an Key Laboratory of Advanced Control and Intelligent Process School of Automation Xi'an University of Posts & Telecommunications Xi'an710061 China Xinjiang University of Political Science and Law Tumushuke832003 China School of Remote Sensing and Information Engineering Wuhan University Wuhan430079 China Department of Computer City University of Hong Kong 999077 Hong Kong School of Communications and Information Engineering Xi'an University of Posts & Telecommunications Xi'an710061 China 

出 版 物：《IEEE Internet of Things Journal》 (IEEE Internet Things J.)

年 卷 期：2025年

核心收录：

学科分类：0810[工学-信息与通信工程] 0808[工学-电气工程] 08[工学] 0802[工学-机械工程] 0835[工学-软件工程] 0825[工学-航空宇航科学与技术] 0701[理学-数学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

主　　题：Unmanned aerial vehicles (UAV) 

摘      要：With the proliferation of IoT devices, there is an escalating demand for enhanced computing and communication capabilities. Mobile Edge Computing (MEC) addresses this need by relocating computing resources to the network edge, thereby delivering swifter and more efficient services. This paper introduces a computation offloading and energy consumption optimization framework that leverages Reconfigurable Intelligent Surfaces (RIS), Unmanned Aerial Vehicles (UAVs), and MEC. The scheme aims to maximize energy efficiency through the optimization of task allocation, RIS phase shifts, and UAV trajectories. By employing the Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MATD3) reinforcement learning algorithm, the paper further refines UAV trajectories and RIS configurations. The simulation results indicate that the proposed method surpasses the traditional Concave-Convex Procedure (CCCP) algorithm in both UAV trajectory control and RIS configuration, demonstrating quicker convergence and enhanced stability. The method proves to be adaptable to diverse environments and tasks, showcasing notable benefits in RIS-assisted interference suppression, particularly with large RIS, thereby enhancing UAV data reception rates. Additionally, MATD3 exhibits faster and smoother convergence for extended task durations and smaller RIS scenarios. Simulation results reveal that UAVs tend to move closer to RIS, with energy efficiency falling as IoT tasks increase, affirming the proposed algorithm s high energy efficiency and effectiveness. © 2025 IEEE.