移动场景中基于DQN的毫米波MISO系统下行链路波束成形  

DQN based downlink-beamforming for millimeter wave MISO system in mobile environment

作　　者：李中捷[1] 吴园君 金闪 钟小辉 LI Zhongjie;Wu Yuanjun;JIN Shan;ZHONG Xiaohui(College of Electronic and Information Engineering & Hubei Key Laboratory of Intelligent Wireless Communications,South-Central University for Nationalities,Wuhan 430074,China)

机构地区：[1]中南民族大学电子信息工程学院&智能无线通信湖北重点实验室,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2021年第3期278-285,共8页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：国家自然科学基金资助项目(61379028,61671483);湖北省自然科学基金资助项目(2016CFA089);中央高校基本科研业务费专项资金资助项目(CZY19003)。

摘　　要：针对移动场景中毫米波MISO系统的可靠性通信问题,提出一种基于深度Q网络(deep Q network,DQN)的下行链路波束成形优化方案.考虑用户快速移动导致的多普勒频移来构建移动场景中的毫米波信道模型,结合功率控制和干扰约束,将MISO系统下行链路的波束成形联合设计方案(joint beamforming-power control and interference coordination,JB-PCIC)建模为非凸优化问题.根据最大化信干噪比(signal to interference plus noise ratio,SINR)及用户的位置坐标,构建智能体-无线环境交互,利用DQN的贪婪属性估计动作的未来奖励,求解该优化问题.仿真结果表明:与穷举搜索的蛮力算法相比,所提出的基于DQN算法的JB-PCIC方案在显著降低复杂度的情况下,可以达到蛮力算法的速率和容量.In order to address the reliability communication problem of mmWave MISO system in mobile environment,an optimization approach for downlink beamforming based on deep Q network is proposed.Fast mobile mmWave channel model according to the Doppler frequency shift caused by the movement of users is constructed,and then the joint design of beamforming,power control,and interference coordination is formulated as a non-convex optimization problem in MISO downlink system.According to the signal to interference plus noise ratio(SINR)and the user′s location coordinates,the agent-environment interaction is constructed,and by using the greedy nature of DQN to estimate the future reward of the action to address the optimization problem.The simulation indicated that with respect to exhaustive search algorithm-Brute Force,the proposed JB-PCIC algorithm based on DQN can achieve comparable performance while operating at much lower complexity.

关 键 词：波束成形 JB-PCIC方案 DQN网络 信干噪比 速率和容量 

分 类 号：TN929.5[电子电信—通信与信息系统]