Joint Optimization of Offloading, Phase-Shift, Trajectory and Resource Allocation in Priority-Aware IRS-Assisted UAV-MEC Systems

Abstract

Unmanned aerial vehicle (UAV)-assisted mobile edge computing (MEC) utilizes low-altitude resources to provide computation services for ground users (GUs). However, the wireless channels connecting UAVs and GUs are often weak, and the priority of heterogeneous tasks is usually ignored, leading to unsatisfactory quality of service (QoS). To overcome this challenge, this paper considers task priority and integrates intelligent reflecting surface (IRS) into traditional UAV-MEC to enhance offloading efficiency. Firstly, we present a priority-aware multi-UAV single IRS-aided MEC system, where GUs can process the tasks with different priorities and tolerable delays locally or offload them to UAV via IRS. Subsequently, we formulate the problem as a joint optimization of task offloading, IRS phase shift, UAV trajectory and resource allocation, with the objective of maximizing the weighted sum of time and energy efficiency. To tackle this complicated Mixed Integer Non-linear Programming (MINLP) problem, we decouple it into three sub-problems and design corresponding algorithms to solve them: 1) we propose a Dynamic Dual-Weighting (DDW)-based Offloading Algorithm (DDWOA) to determine GUs’ offloading strategy; 2) we propose a Genetic Algorithm-based IRS Phase-shift Optimization method (GAIPO) to maximize selected GUs’ uplink channel gain; 3) we devise a Multi-Agent Proximal Policy Optimization (MAPPO)-DDW based UAV Trajectory and Resource Allocation approach (MDUTRA) to maximize system utility. Finally, extensive experimental results indicate that the system utility of our proposed method is dramatically higher than other benchmark schemes under different scenes.