面向车联网的DQN多级边缘缓存算法  

DQN multi-level edge caching algorithm for Internet of vehicles

作　　者：李胜 韩龙哲 敖晨晨 汪泽恒 张波 张翼英[3] LI Sheng;HAN Longzhe;AO Chenchen;WANG Zeheng;ZHANG Bo;ZHANG Yiying(School of Information Engineering,Nanchang Institute of Technology,Nanchang 330099,China;State Grid Smart Grid Research Institute Co.Ltd,Beijing 102211,China;College of Artificial Intelligence,Tianjin University of Science&Technology,Tianjin 300222,China)

机构地区：[1]南昌工程学院信息工程学院,江西南昌330099 [2]国网智能电网研究院有限公司,北京102211 [3]天津科技大学人工智能学院,天津300222

出　　处：《南昌工程学院学报》2022年第6期78-85,共8页Journal of Nanchang Institute of Technology

基　　金：江西省教育厅科学技术研究项目(GJJ190957);国家自然科学基金资助项目(61962036);教育部留学回国人员科研启动基金项目(教外司留[2015]1098号)。

摘　　要：针对智慧交通和车载网络应用爆发式增长及车辆用户海量数据高效传输的迫切需求,提出了基于“车辆—边缘云—中心云”三级车联网协作边缘缓存模型,以实现网络缓存资源的自适应分配。基于该模型,对路侧单元的缓存交付与内容替换过程进行建模,设计了降低内容交付时延的联合缓存优化算法;鉴于车联网的动态性、随机性和时变性,构建了基于深度Q学习的多级边缘缓存算法,求解最优缓存策略。实验结果表明,在不同网络环境参数下,所提出的算法在缓存命中率方面获得了明显提升。In response to the explosive growth of smart transportation and in-vehicle network applications and the urgent need for efficient transmission of massive data from vehicle users,a three-level vehicle-edge cloud-central cloud edge cache model is proposed to realize the adaptive allocation of network cache resources.Based on this model,the roadside unit’s cache delivery and content replacement process are modeled,and a joint cache optimization algorithm is designed to reduce the delay of content delivery.In view of the dynamic,random and time-varying nature of the Internet of vehicles,a DQN-based reinforcement learning algorithm is proposed to solve the optimal caching strategy.Experimental results show that under different network environment parameters,the proposed algorithm has achieved a significant performance improvement in cache hit rate.

关 键 词：车联网 边缘计算 深度强化学习 内容缓存 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]