一种任务驱动的车联网边缘卸载策略  

A Task-driven Edge Offloading Strategy for Internet of Vehicles

作　　者：赵晓焱[1] 高源志 张俊娜[1] 袁培燕[1] ZHAO Xiaoyan;GAO Yuanzhi;ZHANG Junna;YUAN Peiyan(School of Computer and Information Engineering,Henan Normal University,Xinxiang 453007,China)

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007

出　　处：《郑州大学学报（理学版）》2024年第4期34-40,共7页Journal of Zhengzhou University:Natural Science Edition

基　　金：国家自然科学基金项目(62072159,U1804164,61902112);河南省科技攻关项目(222102210011);河南省高等学校重点项目(19A510015,20A520019,20A520020)。

摘　　要：边缘计算为解决未来车联网中移动流量的爆炸式增长提供了可行范式,然而位置的动态变化以及计算任务的多样性和差异性,使得资源有限的边缘服务器很难在规定时间内完成区域内多车辆任务的并行处理需求。基于此,以最小化时延为目标,提出一种结合深度确定性策略梯度算法的任务驱动卸载策略。首先,结合差异性任务类型和紧迫程度进行预处理,构建了一种基于最大延迟容忍度的任务动态优先级调整模型;然后,利用道路区域内的车辆拓扑和通信半径,提出了基于网络密度和负载均衡的动态协作簇划分方法,解决了多样性任务的动态协作卸载优化问题。实验结果表明,所提算法在收敛性、卸载时延及卸载命中率等方面具有性能优势。Edge computing provided a feasible paradigm to address the explosive growth of mobile traffic in the future Internet of Vehicles.However,it was difficult for resource-limited edge servers to complete the parallel processing requirements of multi-vehicle tasks in a region within a specified time due to the diversity and variability of computing tasks and the dynamic location change.Therefore,with the goal of minimizing latency,a task-driven offloading strategy combined with a deep deterministic policy gradient algorithm was proposed.Firstly,a dynamic task priority adjustment model based on the maximum delay tolerance was constructed by preprocessing the different task types and urgency.Then,a dynamic cooperative cluster partition method based on network density and load balancing was proposed by using the vehicle topology and communication radius in a road region,which solved the dynamic cooperative offloading optimization problem of diverse tasks.Experimental results showed that the proposed algorithm had performance advantages in convergence,offloading delay and offloading hit rate.

关 键 词：车联网 边缘计算 任务卸载 协作 动态优先级 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]