基于马氏决策的5G网络切片虚拟资源分配研究  

Research on Virtual Resource Allocation of 5GNetwork Slice Based on Markov Decision

作　　者：曾坚毅 许济金[2] ZENG Jian-yi;XU Ji-jin(Fuzhou Institute of Technology,Fuzhou Fujian 350506,China;College of mechanical and electrical engineering,Fujian agriculture and Forestry University,Fuzhou Fujian 350002,China)

机构地区：[1]福州理工学院,福建福州350506 [2]福建农林大学机电工程学院,福建福州350002

出　　处：《计算机仿真》2022年第6期242-246,共5页Computer Simulation

基　　金：2021年移动通信和物联网福建省高校工程研究中心开放课题(KFKT2021001);2021年福州理工学院科技类科研基金重点项目(FTKY21040);2019年福州市科技计划项目(2019-SG-10)。

摘　　要：为了解决5G网络资源的分配问题,提高网络虚拟资源的利用率,为用户定制最佳的虚拟资源分配方案,提出基于马氏决策的5G网络切片虚拟资源分配方法。首先根据检测的切片信息对虚拟资源进行分配,将所有用户分配到不同的资源模块中。然后通过联络特征分析方法和特征空间重组技术对虚拟资源分配的特征集和结构进行构建,对虚拟资源进行优化设计,建立网络信道的均衡模型,得出5G网络切片虚拟资源的调度模型。最后结合贝尔曼方程及马氏决策过程的动态规划迭代算法求解出运营商最优收益解和网络资源有效利用率。实验结果表明,基于马氏决策的5G网络切片虚拟资源分配方法无论是在网络运营商的长期累积收益方面,还是在用户切片请求阻塞概率方面均优于传统方法,有效验证了所提策略的合理性与准确性。In order to solve the problem of 5 G network resource allocation, improve the utilization of network virtual resources, and customize the best virtual resource allocation scheme for users, a 5 G network slice virtual resource allocation method based on the Markov decision process is proposed. Firstly, virtual resources were allocated according to the detected slice information, and all users were allocated to different resource modules. Then, the feature set and structure of virtual resource allocation were constructed by the method of contact feature analysis and feature space reorganization, and the virtual resource was optimized. The equilibrium model of the network channel was established, and the scheduling model of the 5 G network slice virtual resource was obtained. Finally, the optimal revenue solution and the effective utilization of network resources were obtained by combining the Bellman equation and the dynamic programming iterative algorithm of the Markov decision process. The experimental results show that the 5 G network slice virtual resource allocation method based on the Markov decision process is superior to the traditional method in terms of both the long-term cumulative revenue of network operators and the blocking probability of user slice requests, which verifies the rationality and accuracy of the proposed strategy.

关 键 词：网络资源 虚拟资源分配 调度模型 贝尔曼方程 马氏决策 

分 类 号：TN929.5[电子电信—通信与信息系统]