.png){kind=icon}
.png){kind=icon}
Return
Joint Optimization of UAV-Mounted RIS-Assisted Mobile Edge Computing Using Deep Reinforcement Learning
.png){kind=icon}
PRE
AI
DOI：10.1007/s10922-026-10098-7.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Mobile Edge Computing (MEC) is an emerging technology with great potential in mobile user networks, enabling terminal devices to execute computation-intensive tasks locally or offload them to nearby edge servers. However, the communication link quality during the offloading process can be severely degraded due to obstacles such as buildings. To address this challenge, we introduce a Unmanned Aerial Vehicle–mounted Reconfigurable Intelligent Surface (UAV-RIS), which provides alternative communication paths in the air to enhance the connectivity between terminal devices and edge servers. By dynamically adjusting the phase configuration of the RIS, the overall performance of the MEC system can be significantly improved. In this study, we consider a UAV-RIS-assisted MEC system and propose a joint optimization scheme for local computation, task offloading, and RIS phase control, while taking into account stochastic task arrivals and channel variations. To solve this optimization problem, we propose an innovative Deep Reinforcement Learning (DRL) framework, in which the RIS phase control is optimized using the Deep Deterministic Policy Gradient (DDPG) algorithm, while user power allocation is optimized through a multi-user parallel TD3 algorithm. Simulation results demonstrate that the proposed joint optimization framework outperforms conventional centralized DDPG, TD3, and MADDPG schemes. Specifically, under high-load conditions, it reduces total device energy consumption by up to 85.9% (compared to DDPG) and decreases the average task queue length by 26.7% (compared to TD3), thereby achieving a superior trade-off between energy efficiency and task processing delay.
Keywords:
Mobile edge computing
Reconfigurable intelligent surface
Unmanned aerial vehicle
Deep reinforcement learning
Phase control
Power allocation
Journal
IF:
3.9
Papers:
1.0K
Citations:
1.3K
Organization
View more organizations