摘要
以最小化卸载成本为目标,提出一种结合轨迹预测的任务卸载策略,将任务卸载转化为服务器节点选择问题。构建一种基于时间序列的车辆移动轨迹预测模型,将其表述为一个非线性回归任务;依据车辆位置信息与通信范围,提出一种基于最短通信距离的动态协作簇建立方法,利用服务器计算能力和传输成本均衡边缘网络负载,减少车辆移动形成的系统开销;利用马尔可夫决策过程,结合移动轨迹预测和动态边缘服务器簇设计任务卸载策略,解决多边缘服务器覆盖场景下的服务器选择问题。试验结果表明,所提算法与其他算法相比,任务卸载成本在简单与复杂移动轨迹下至少降低了80%和57.8%,有效减少多边缘服务器协作时的轨迹预测误差和成本开销。
A task offloading strategy combined with trajectory prediction was proposed to minimize the offloading cost,and the task offloading was transformed into a server node selection problem.A time-series-based vehicle movement trajectory prediction model was constructed and presented as a nonlinear regression task.According to the vehicle location information and communication range,a dynamic edge cluster method was proposed based on the shortest communication distance.Server computing power and transmission cost were utilized to optimize the load distribution in the edge network and reduce system overhead caused by vehicle movement.The server selection problem in a multi-edge server coverage scenario was effectively addressed by designing a task offloading strategy based on moving trajectory prediction and dynamic edge server clusters using Markov decision process.Experimental results showed that compared with other algorithms,the proposed algorithm could reduce the task offloading cost by 80% and 57.8% at least on simple and complex movement trajectory.The trajectory prediction error and cost of multi-edge server collaboration could be effectively reduced.
作者
赵晓焱
高源志
张佳乐
张俊娜
袁培燕
ZHAO Xiaoyan;GAO Yuanzhi;ZHANG Jiale;ZHANG Junna;YUAN Peiyan(College of Computer and Information Engineering,Henan Normal University,Xinxiang 453007,Henan,China;Engineering Lab of Intelligence Business&Internet of Things(Henan Normal University),Xinxiang 453007,Henan,China)
出处
《山东大学学报（工学版）》
CAS
CSCD
北大核心
2024年第1期52-62,共11页
Journal of Shandong University（Engineering Science）
基金
国家自然科学基金资助项目(62072159)
河南省科技攻关资助项目(222102210011、232102211061)。
关键词
车联网
边缘计算
任务卸载
马尔可夫决策
轨迹预测
internet of vehicles
edge computing
task offloading
Markov decision
trajectory prediction
