基于用户位置预测的移动边缘缓存算法  

Mobile Edge Caching Algorithm Based on User Location Prediction

作　　者：陈晓奇 朱琦[1] CHEN Xiaoqi;ZHU Qi(Jiangsu Key Laboratory of Wireless Communication,Nanjing University of Posts and Telecommunication,Nanjing,Jiangsu 210000,China)

机构地区：[1]南京邮电大学江苏省无线通信重点实验室,江苏南京210000

出　　处：《信号处理》2023年第5期929-937,共9页Journal of Signal Processing

基　　金：国家自然科学基金(61971239,92067201);江苏省重点研发计划(BE2022068-2)。

摘　　要：物联网和5G技术的发展带来了移动流量的大量增长,这极大加重了回程链路的负担。移动边缘缓存技术可以将一部分文件缓存在基站中,避免了回程链路上内容的重复传输,有效缓解了回程链路的负担。为了提升运营商的缓存收益,本文提出了一种基于用户位置预测的多基站缓存算法。该算法采用LSTM(Long Short-Term Memory)模型,通过大量真实数据对模型进行训练,对每一位用户的位置进行预测,进而得到每一时刻基站服务范围内的用户群体。基于预测得到的信息,结合用户偏好信息,预测得到文件请求分布。在基站缓存容量有限的约束下,结合文件请求分布,构建了运营商收益最大化问题,通过动态规划算法求得全局最优缓存策略。仿真结果表明,本文算法可以有效降低回程链路开销,提高运营商的缓存收益。The explosive growth of mobile traffic brought about by the development of the Internet of Things and 5G technology has greatly increased the burden of backhaul.Mobile edge caching technology can cache some files in the base station,which can avoid the repeated transmission of the content,and effectively relieve the burden of the backhaul.In order to improve the operator’s cache revenue,this paper proposes a multi-base station cache algorithm based on user position prediction.The algorithm uses the Long Short-Term Memory(LSTM)model to predict the user location.The model is trained by a large amount of real dataset,and the location information of each user is predicted.Then the user group within the service range of the base station at every moment is obtained.Based on the predicted information,combined with user preference information,the file request distribution is predicted.Under the constraint of limited base station cache capacity,the problem of maximizing the operator’s revenue is constructed.Combined with the file request distribution information,the global optimal cache strategy is obtained by dynamic programming algorithm.Simulation results show that the proposed algorithm can effectively reduce the backhaul cost and improve operator’s cache revenue.

关 键 词：移动边缘缓存 用户移动预测 动态规划 

分 类 号：TN929.5[电子电信—通信与信息系统]