面向高移动性车联网场景的V2X卸载决策算法OA
Predictive V2X offloading decision algorithm for high mobility vehicular network scenarios
针对V2X场景中计算资源不足、任务卸载不合理导致的高时延和能耗问题,提出一种在车辆与其他通信设备(vehicle-to-everything,V2X)场景中多节点协同并行计算的分布式卸载策略.设计了一个云-边-端-车的 4 层卸载架构,结合长短期记忆(long short-term memory,LSTM)网络与卡尔曼滤波构建车辆位置预测模型,为任务车辆提供可卸载的协同节点,使用改进的Q-learning算法实现资源的最优分配.通过对比多种卸载方案的数据表明,所提算法任务卸载的时延与能耗的加权和降低了约 11.4%.
Aiming at the problem of insufficient computing resources of road side units(RSU)in offloading hotspots,we propose a new V2X offloading decision algorithm.Firstly,the computing model of local and surrounding vehicle resources,edge server and cloud server is constructed.Based on the constraints such as maximum tolerance delay and available resources,the offloading mode of tasks is pre-determined.Secondly,according to the offloading mode of edge server and V2V,the vehicle position prediction model is constructed by combining the long-short term memory(LSTM)network and Kalman filter,and the set of edge server and removable vehicle can be generated for offloading.Finally,Q-learning algo-rithm is used to achieve the optimal allocation of uninstallation tasks among multiple nodes.Simulation results demonstrate that the proposed algorithm significantly reduces the weighted sum of offloading latency and energy consumption by approxi-mately 11.4%.
彭维平;蒋崟梦;王戈;宋成
河南理工大学 计算机科学与技术学院,河南 焦作 454003河南理工大学 计算机科学与技术学院,河南 焦作 454003河南理工大学 计算机科学与技术学院,河南 焦作 454003河南理工大学 计算机科学与技术学院,河南 焦作 454003
信息技术与安全科学
车联网边缘计算卸载位置预测长短期记忆(LSTM)网络卡尔曼滤波强化学习
vehicular networksedge computing offloadinglocation predictionlong short-term memory(LSTM)networkKalman filteringreinforcement learning
《重庆邮电大学学报（自然科学版）》 2026 (1)
20-29,10
国家自然科学基金项目(62162009) National Natural Science Foundation of China(62162009)
评论