缓存辅助的MEC卸载和资源分配研究  

Research on cache-assisted MEC offloading and resource allocation

作　　者：朱绍恩 ZHU Shao′en(School of Communications and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学通信与信息工程学院,南京210003

出　　处：《智能计算机与应用》2024年第10期149-157,共9页Intelligent Computer and Applications

基　　金：江苏省重点研发计划(BE2020084-3,BE2021013-2)。

摘　　要：由于MEC在面对大规模数据传输和计算需求时存在性能瓶颈,在实际应用中,通过将任务数据缓存到本地,可以大大减少数据传输时间和网络带宽的消耗,提高MEC的性能。本文研究了多边缘服务器多移动用户缓存辅助场景下的系统长期平均开销优化问题,提出了一种计算卸载、资源分配和缓存决策联合优化方案。该方案基于D3QN框架,利用了遗传算法和KKT分别获得本地和边缘计算资源分配,基于任务请求概率分布更新MEC服务器缓存空间,并通过D3QN网络的学习和保序量化得到近似最优的卸载决策。仿真结果表明,在不同系统参数下,本文所提出方案相较于其他方案具有更佳的性能。Due to the performance bottleneck of MEC in facing large-scale data transmission and computing demands,caching task data locally can significantly reduce data transfer time and network bandwidth consumption,thereby improving the performance of MEC in practical applications.This paper investigates the optimization problem of long-term average system cost in a scenario of multiple edge servers and multiple mobile users with cache assistance,and proposes a joint optimization scheme for computing offloading,resource allocation,and caching decision-making.The proposed scheme is based on the D3QN framework and utilizes genetic algorithm and KKT to respectively obtain local and edge computing resource allocation.And the scheme updates the MEC server cache space based on the probability distribution of task requests and obtains an approximately optimal offloading decision through learning and quantization of the D3QN network.Simulation results show that the proposed scheme has better performance compared with other schemes under different system parameters.

关 键 词：移动边缘计算 深度强化学习 计算卸载 资源分配 数据缓存 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]