基于毫米波MIMO的车车通信混合预编码研究  

Research on hybrid precoding for T2T communication based on mm Wave MIMO

作　　者：杨宗烨 吴卫[1] 董昱[1] YANG Zongye;WU Wei;DONG Yu(School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070

出　　处：《铁道科学与工程学报》2023年第12期4484-4492,共9页Journal of Railway Science and Engineering

基　　金：轨道交通工程信息化国家重点实验室资助项目(SKLKZ22-06);甘肃省科技计划资助项目(20YF8GA037)。

摘　　要：高速铁路普遍采用传统的车地(Train-to-Ground,T2G)通信,针对传统T2G通信存在轨旁设备多、传输延时高和系统容量不足的问题,提出一种车车(Train-to-Train,T2T)通信混合预编码(Hybrid Precoding,HP)的传输方案来弥补T2G通信的不足。在T2T通信中引入毫米波(Millimeter Wave,mmWave)、大规模多输入多输出(Multiple Input Multiple Output,MIMO)和HP等第5代移动通信(Fifth Generation Mobile Communication,5G)关键技术,并重新设计一种适用于HP的等效全连接神经网络模型,同时结合反向传播(Back Propagation,BP)算法和自适应增量(Adaptive Delta,AD)算法,进行基于毫米波大规模MIMO的T2T通信HP研究。研究结果表明:所给训练集越大,AD-BP算法的拟合情况就越好,理想输出和实际输出的误差越小,经过多次迭代训练最终收敛于频谱效率。采用等效网络算法的T2T通信HP和最优无约束预编码相比,二者性能相近。同时,对比传统HP常采用的迫零(Zero Forcing,ZF)算法和最小均方误差(Minimum Mean Square Error,MMSE)算法,采用等效网络算法的T2T通信HP的接收信噪比(Signal-to-Noise Ratio,SNR)能满足T2T传输的最低要求且频谱效率更优。在固定数据流的情况下,提高射频(Radio Frequency,RF)链的数量所带来系统性能的提升有限且获益不高,但却导致系统的复杂度提高,所以要在不降低系统性能的情况下尽可能减少RF链数量来降低系统的复杂度。研究结果可为未来实际设计高速铁路T2T通信传输提供理论指导。High-speed railway generally adopts traditional Train-to-Ground(T2G)communication.Traditional T2G communication has the problems of much trackside equipment,high transmission delay and insufficient system capacity;therefore,a Train-to-Train(T2T)communication Hybrid Precoding(HP)transmission scheme can be proposed to compensate for the shortcomings of T2G communication.Millimeter Wave(mmWave),Multiple Input Multiple Output(MIMO)and HP into T2T communication and other critical technologies of 5th Generation Mobile Communication(5G)were introduced,an equivalent fully connected neural network model suitable for HP was designed,and the Back Propagation(BP)algorithm and the Adaptive Delta(AD)algorithm were combined to conduct T2T communication HP research based on mmWave massive MIMO.The results show that the larger the training set,the better the fitting of the AD-BP algorithm,and the closer the error between the ideal output and the actual output.The training result finally converges to the spectral efficiency after multiple changes in training.Compared with the optimal unconstrained precoding of T2T communication HP using equivalent network algorithms,the performance of the two is similar.At the same time,compared with the Zero Forcing(ZF)algorithm and the Minimum Mean Square Error(MMSE)algorithm commonly used by traditional HP,the Signal-to-Noise Ratio(SNR)of T2T communication HP using an equivalent network algorithm can meet the minimum requirements of T2T transmission and have better spectral efficiency.In the case of fixed data streams,increasing the number of Radio Frequency(RF)chains brings limited and low benefits from system performance.Still,it leads to increased system complexity,so reducing the system’s complexity by reducing the number of RF chains as much as possible without reducing system performance is necessary.The research results can provide theoretical guidance in the future practical design of high-speed railway T2T communication transmission.

关 键 词：T2T通信 毫米波 多输入多输出 混合预编码 

分 类 号：U285[交通运输工程—交通信息工程及控制]