M2M通信中基于多智能体强化学习的无线资源分配算法  

Multi-agent reinforcement learning based resource allocation for M2M communication

作　　者：徐少毅[1] 郑姗姗 XU Shaoyi;ZHENG Shanshan(School of Electronic and Information Engineering,Bcijing Jiaotong University,Beijing 100041,China)

机构地区：[1]北京交通大学电子信息工程学院,北京100044

出　　处：《北京交通大学学报》2018年第5期1-9,共9页JOURNAL OF BEIJING JIAOTONG UNIVERSITY

基　　金：国家自然科学基金(61571038;61471030);国家科技重大专项(2016ZX03001011-004);中央高校基本科研业务费专项资金(2016JBZ003)~~

摘　　要：蜂窝网络因为其广覆盖、高可靠和支持高速移动等优点,是机器对机器(M2M)通信的理想载体.然而,由于机器类型通信设备具有的业务种类繁多和数量大等特点,现有的无线资源分配算法并不完全适用.为了解决这个问题,不同于传统的集中式资源分配算法,提出了基于多智能体强化学习的分布式无线资源分配算法.具有强化学习能力的机器类通信(MTC)设备可以自主选择资源块和功率等级,以达到在较低的功率消耗下得到较高的用户体验质量的目标.为了实现不同设备之间的协作,引入了多智能体强化学习,每个智能体有能力预测其他智能体的策略.仿真结果表明:本文算法在用户体验质量、功率消耗和计算复杂度方面达到了很好的性能.Cellular network is an ideal carrier for Machine to Machine （M2M） communication benefited from its characteristics such as broad coverage, high reliability and supporting for high speed mobility. However, because of the various service types and huge amount of the Machine Type Communication （MTC） devices, the existing resource allocation algorithm is not fully applicable. In order to deal with this matter, in this paper, different from the traditional centralized algorithm, a distributed wireless resource allocation algorithm based on reinforcement learning is developed where the MTC devices with the ability of reinforcement learning can smartly choose resource blocks and power levels. The goal of them is to increase the Quality of Experience （QoE） with low power consumption. In order to achieve the collaboration between different devices, the multi agent reinforcement learning algorithm is also introduced by which the MTC devices have the ability to predict the other devices＇ activities. The simulation results show that the proposed algorithm achieves good performance in terms of QoE, power consumption and compu tation complexity.

关 键 词：无线通信 机器对机器 强化学习 用户体验质量 资源分配 

分 类 号：TN92[电子电信—通信与信息系统]