车联网中改进粒子群算法的任务卸载策略

Improved particle swarm algorithm for task offloading in vehicular networks

当前,多数车联网任务卸载工作仅考虑时延因素将任务卸载至边缘服务器执行(LOCAL-MEC),但是,车载单元仍有一定的计算能力可以利用。针对上述问题,研究了任务卸载的总代价即时延和能耗两个目标,提出一个将车辆自身的计算单元、附近车辆的计算单元与边缘服务器协同计算的任务卸载模型。该模型既考虑了任务的优先关系,又同时考虑了系统的时延和能耗。通过借鉴模拟退火算法思想并引入压缩因子改进粒子群算法来实现任务卸载。实验结果表明:与其他任务卸载策略相比,提出的任务卸载策略优化效果明显,TPSO算法的总代价为传统粒子群算法的53.8%、LOCAL-MEC策略的27.1%、DCOS(distributed computation offloading scheme)算法的78%,并且适用于多种现实场景。

At present,most existing works focus on offloading tasks to MEC servers,in view of the total process time of the computation tasks.However,the on board unit(OBU)still has its computing capacity.To solve these problems,this research aimed to optimize the time and energy consumption of tasks and proposed a collaborative multi-task computing offloading model,where OBU in the task-solving vehicle and neighbor vehicles compute in collaboration with the MEC servers.This model also considered the energy and time consumption as well as priority among different tasks.It introduced the idea of simulated annealing and the constriction coefficient to improve the particle swarm optimization algorithm and got solution.Experimental results show that compared with other offloading strategies,the proposed task offloading strategy has obvious effects on the total cost optimization,the solutions quality of TPSO is 53.8%of PSO algorithm,27.1%of the LOCAL-MEC strategy and 78%of the DCOS,and it can adapt to various application scenarios.