Robust Design for Computing Uncertainty and Flying Jittering in Multi-UAV-Assisted MEC

Yucheng Gong, Bin Li

Keywords: mobile edge computing, distributionally robust optimization, deep reinforcement learning, robust design, computation uncertainty

TL;DR: Robust computation offloading in multi-UAV MEC networks using DRO-SAC.

Abstract:

This paper investigates the problem of joint computation and communication optimization in multiple unmanned-aerial-vehicles (UAVs)-assisted mobile edge computing (MEC) networks, where the practical challenges arising from task data sizes uncertainty and UAV fly jittering are taken into account.
We propose a robust optimization framework to address these uncertainties, aiming to minimize the weighted energy consumption. This is achieved by jointly optimizing UAV trajectories, task partitioning, and the allocation of computation and communication resources across multiple UAVs.
The problem is solved using a distributionally robust optimization soft actor–critic algorithm, which enhances system robustness by accounting for worst-case task demand distributions.
Numerical simulations demonstrate that the proposed algorithm significantly outperforms conventional deep reinforcement learning approaches in terms of energy consumption, while ensuring reliable task completion in dynamic environments.

Submission Number: 9