Collaboration in the Sky: A Distributed Framework for Task Offloading and Resource Allocation in Multi-Access Edge Computing

TL;DR

This paper proposes a collaborative multi-UAV-assisted MEC system integrated with terrestrial base stations, optimizing task offloading and resource allocation to significantly reduce latency for mobile users.

Contribution

It introduces a novel collaborative framework for UAV-assisted MEC, formulating and solving a complex non-convex problem with decomposition and optimization techniques.

Findings

Latency reduced by up to 40.7% compared to greedy methods.

Effective resource allocation improves system performance.

Collaborative UAV operation enhances MEC system efficiency.

Abstract

Recently, unmanned aerial vehicles (UAVs) assisted multi-access edge computing (MEC) systems emerged as a promising solution for providing computation services to mobile users outside of terrestrial infrastructure coverage. As each UAV operates independently, however, it is challenging to meet the computation demands of the mobile users due to the limited computing capacity at the UAV's MEC server as well as the UAV's energy constraint. Therefore, collaboration among UAVs is needed. In this paper, a collaborative multi-UAV-assisted MEC system integrated with a MEC-enabled terrestrial base station (BS) is proposed. Then, the problem of minimizing the total latency experienced by the mobile users in the proposed system is studied by optimizing the offloading decision as well as the allocation of communication and computing resources while satisfying the energy constraints of both mobile users and UAVs. The proposed problem is shown to be a non-convex, mixed-integer nonlinear problem (MINLP) that is intractable. Therefore, the formulated problem is decomposed into three subproblems: i) users tasks offloading decision problem, ii) communication resource allocation problem and iii) UAV-assisted MEC decision problem. Then, the Lagrangian relaxation and alternating direction method of multipliers (ADMM) methods are applied to solve the decomposed problems, alternatively. Simulation results show that the proposed approach reduces the average latency by up to 40.7\% and 4.3\% compared to the greedy and exhaustive search methods.