| 引用本文： |
-
艾臻,付峰,余健,等.多臂赌博机赋能大规模MIMO终端调度和预编码设计[J].电讯技术，2025，（12）：2103 - 2112. [点击复制]
- AI Zhen,FU Feng,YU Jian,et al.Multi-armed Bandit Empowered High Spectral-efficient Terminal Scheduling and Precoding for 5G Massive MIMO[J].，2025，（12）：2103 - 2112. [点击复制]
|
|
| 摘要: |
| 为提升频分双工大规模多输入多输出（Multiple-Input Multiple-Output，MIMO）系统的频谱效率，提出了基于多臂赌博机（Multi-armed Bandit,MAB）的预编码设计方法。该方法利用MAB从历史数据中获取信道协方差矩阵（Channel Covariance Matrix,CCM）信息进行预编码设计，避免了CCM的估计，同时无需对瞬时信道状态信息估计，有效降低了导频开销并提升了系统频谱效率。首先，利用较少的时隙获取初步的终端路径角度信息，并基于最小反距离和准则，提出了终端调度算法，使终端在移动过程中仍能保持信道的正交，减小了终端间干扰。其次，将各终端预编码向量和接收信号能量分别表征为MAB中的动作和奖励值，根据置信度上界策略，提出了各终端动作集迭代以及确定终端动作的算法。该算法可逐步收敛到使终端接收能量总和最大的预编码向量。仿真结果验证了所提方法的频谱效率较以往方法能够提升5%~20%。 |
| 关键词: 大规模MIMO 终端调度 预编码 多臂赌博机 |
| DOI：10.20079/j.issn.1001-893x.240629003 |
|
| 基金项目: |
|
| Multi-armed Bandit Empowered High Spectral-efficient Terminal Scheduling and Precoding for 5G Massive MIMO |
| AI Zhen,FU Feng,YU Jian,DING Hong |
| (State Grid Electric Power Research InstituteNARI Group Corporation,Nanjing 211106,China) |
| Abstract: |
| In order to improve the spectral efficiency of frequency division duplex massive multiple-input multiple-output(MIMO) systems,the authors propose a precoding design method based on multi-armed bandit(MAB).This method utilizes MAB to obtain channel covariance matrix(CCM) information from historical data,and uses the obtained CCM for precoding design,which avoids the estimation of CCM and does not require estimation of instantaneous channel state information,effectively reducing pilot overhead and improving system spectral efficiency.First,fewer time slots are used to obtain preliminary user path angles,and based on the minimum inverse distance sum and criterion,a terminal scheduling algorithm that enables terminals to maintain channel orthogonality during movement is proposed,reducing inter-terminal interference.Second,the precoding vectors and received signal energy of each terminal are represented as the action and reward in the MAB,respectively.According to the upper confidence bound strategy,an algorithm for iterating and determining terminal actions for each terminal action set is proposed.This algorithm can gradually converge to the precoding vector that maximizes the total energy received by the terminal.The simulation results validate that the spectral efficiency of the proposed method can be improved by 5% to 20% compared with that of existing methods. |
| Key words: massive MIMO user scheduling precoding multi-armed bandit |
《电讯技术》编辑部 