异构边缘云架构下的多任务卸载算法
Multitask offloading algorithm under heterogeneous edge cloud architecture
尼俊红 1臧云2
作者信息
- 1. 华北电力大学(保定) 电子与通信工程系,河北 保定 071003;华北电力大学 河北省电力物联网技术重点实验室,河北保定 071003
- 2. 华北电力大学(保定) 电子与通信工程系,河北 保定 071003
- 折叠
摘要
为在资源有限的终端设备上运行计算密集型与时延敏感型应用,同时降低系统时延和能耗,构建边缘云异构网络模型.本文提出了一种H-PSOGA多任务卸载优化算法,并通过无人机、路边单元、车辆等边缘设备以及边缘云服务器进行多任务计算卸载.该算法以先串行再并行的方式将粒子群和遗传算法结合在一起,通过适应度值排序、种群选择、多点交叉、反向变异等操作,利用遗传算法对粒子群进行优选,弥补粒子群算法早熟收敛、陷入局部最优的缺陷.6 种标准测试函数的测试分析以及与基线方案进行仿真对比的结果表明:在用户数较多时,混合优化算法的系统平均开销可降低 26%～43%,可以有效提高收敛精度.
Abstract
To run computing-intensive and delay-sensitive applications on terminal devices with limited resources and to reduce system time delay and energy consumption,an edge cloud heterogeneous network model is construc-ted.In addition,a hierarchical particle swarm optimization with genetic algorithm(H-PSOGA)for multitasking of-floading optimization is proposed for multitasking computational offloading through edge devices,such as drones,roadside units,and vehicles,as well as edge cloud servers.The H-PSOGA combines particle swarm and genetic algorithms in a serial and then parallel manner.The genetic algorithm is used to optimize the particle swarm through operations such as fitness value sequencing calculation,population selection,multipoint crossover,and reverse mu-tation to compensate for the defects in the particle swarm algorithm.These issues include premature convergence and local optima.The test analysis of six standard test functions and the result of simulation comparison with the baseline scheme show that with a large number of users,the average cost can be reduced by 26%to 43%,H-PSOGA can effectively improve convergence accuracy reduce system overhead.
关键词
移动边缘计算/异构网络/边缘节点/任务卸载/粒子群算法/遗传算法/多目标优化/标准测试函数Key words
mobile edge computing/heterogeneous network/edge node/task offloading/particle swarm algorithm/genetic algorithm/multiobjective optimization/standard test function引用本文复制引用
基金项目
国家自然科学基金(61771195)
河北省自然科学基金(F2018502047)
出版年
2024
国家科技期刊平台
NSTL
维普
万方数据