基于无人机的移动边缘计算任务卸载  

A task offloading scheme of mobile edge computing for UAVs

作　　者：刘建华 林柯蒙 衡振宇 刘佳嘉 谢家雨[1] LIU Jianhua;LIN Kemeng;HENG Zhenyu;LIU Jiajia;XIE Jiayu(School of Avionics and Electronics,Civil Aviation Flight University of China,Guanghan 618300,China)

机构地区：[1]中国民用航空飞行学院航空电子电气学院,四川广汉618300

出　　处：《南京邮电大学学报（自然科学版）》2023年第2期36-45,共10页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：四川省科技厅项目(2022JDKP0093,2022JDRC0076);中央高校基本科研业务费基金(ZHIMH2022-004,J2022-025);中飞院科研项目(ZX2021-03)资助项目。

摘　　要：无人机具有高移动性,能帮助移动用户在基础通信设施缺乏的环境中快速部署边缘云。为降低基于无人机的边缘计算成本,提高能量利用效率,提出了一种移动边缘计算(Mobile Edge Computing, MEC)系统的任务卸载方案。首先基于排队论模型构建了多用户边缘计算模型,引入等待概率构造了移动边缘计算系统的成本函数,然后采用权重系数将多目标优化问题转化为求解卸载策略的单目标非线性规划问题,最后采用蒙特卡洛剪枝算法进行求解,并引入决策树剪枝算法降低了算法的复杂度。实验结果表明,文中提出的任务卸载方案在CPU性能较低、任务到达率较大的情况下均能降低成本开销,为基于无人机的边缘计算提供了一种低成本的解决方案。Thanks to the high mobility, unmanned aerial vehicles (UAVs) can quickly help mobile usersdeploy edge cloud in the environment that lacks basic communication facilities. In order to reduce UAVs􀆳cost of edge computing and improve their energy utilization efficiency, we propose a task offloadingscheme of mobile edge computing (MEC). First, a multi⁃user edge computing model based on thequeuing theory model is constructed, and the cost function of the MEC system is constructed byintroducing the waiting probability. Second, the multi⁃objective optimization problem is transformed into asingle objective nonlinear programming problem for solving the offloading problem through weightcoefficients. Finally, the Monte Carlo pruning algorithm is used to solve the problem. Specially, thedecision pruning algorithm is introduced to reduce the computational complexity. The experimental resultsshow that compared with traditional schemes, the proposed offloading scheme can effectively reduce thecost with a low CPU performance and a high task arrival rate, and provide a low⁃cost solution of edgecomputing for UAVs.

关 键 词：无人机 移动边缘计算 蒙特卡洛树搜索算法 决策树剪枝 

分 类 号：TN929.53[电子电信—通信与信息系统]