一种面向移动边缘计算的无人机基站部署方法

UAV Base Station Deployment Method for Mobile Edge Computing

在移动边缘计算(Mobile Edge Computing,MEC)中,本地设备可以将任务卸载到边缘服务器执行,以此来提高服务质量(Quality of Service,QoS)。但在受灾地区或遇到紧急情况时,地面固定的基站可能会出现大面积瘫痪,为了应急通信,无人机(Unmanned Aerial Vehicle,UAV)支持的移动边缘计算系统应运而生。作为新兴的应急通信手段,无人机可以携带边缘服务器,地面用户设备可以将其计算任务卸载给无人机执行,但在多用户网络中部署多个无人机基站是具有挑战性的。为此,重点研究无人机基站的战略部署问题,将该问题建模为多目标优化问题,旨在平衡无人机基站之间的工作负载,最小化地面用户和无人机基站之间的访问延迟。与单目标优化问题相比,多目标之间相互作用并且解不唯一,给模型求解带来了一定困难。为此,提出基于K-中心点(K-medoids)的帕累托边界搜索算法求解该问题,之后进一步提出利用主成分分析算法(Principal Component Analysis,PCA)从帕累托边界中寻找最合适的解作为最终的无人机基站部署策略。实验使用真实的数据集,并与其他几种基线方法进行性能比较,验证了所提解决方案的有效性。

In mobile edge computing(MEC),local devices can offload tasks to edge servers for execution to improve the quality of service(QoS).However,in disaster-stricken areas or in emergencies,ground-fixed base stations may be paralyzed on a large scale.For emergency communications,mobile edge computing systems supported by unmanned aerial vehicles(UAV)have emerged.As an emerging means of emergency communication,drones can carry edge servers,and ground user equipment can offload their computing tasks to the drones for execution.However,it is challenging to deploy multiple UAV base stations in a multi-user network.To this end,this paper focuses on the strategic deployment of UAV base stations,modeling the problem as a multi-objective optimization problem,which aims to balance the workload among UAV base stations and minimize the access delay between ground users and UAV base stations.Compared with single-objective optimization problems,multi-objectives interact with each other and the solutions are not unique,which brings certain difficulties to model solving.For this reason,this paper proposes a Pareto boundary search algorithm based on K-medoids to solve the problem,and then further proposes to use the principal component analysis algorithm(PCA)to find the most suitable solution from the Pareto boundary as the final deployment strategy for UAV base stations.The experiment in this paper uses real data sets and compares the performance with several other baseline methods to verify the effectiveness of the proposed solution.