基于遗传算法的毫米波大规模MIMO系统混合预编码

Genetic Algorithm Based Hybrid Precoding for Millimeter Wave Massive MIMO Systems

目前混合预编码方案中,大多采取高精度的移相器作为模拟预编码的设计基础,使得系统成本增加。针对这一问题,探讨了有限精度射频前端的混合预编码设计。为了实现更高的频谱利用率,考虑到天线权值的最优组合为一NP(Non-deterministic Polynomial)问题,受机器学习启发,采用遗传算法对阵列中阵元的相位取值进行建模设计模拟预编码。通过信道矩阵与模拟预编码矩阵的乘积引入等效信道矩阵,考虑用户间干扰,以最大信干噪比准则进行数字预编码设计。仿真结果表明,该方案得到的混合预编码矩阵其系统性能可逼近全数字预编码矩阵的性能。

In current hybrid precoding schemes,most of the high-precision phase shifters are used as the design basis of the analog precoding,so that the system cost increases.For this problem,the hybrid precoding design of limited precision radio frequency(RF) front-end is discussed.In order to achieve higher spectrum utilization,considering that the optimal combination of antenna weights is a non-deterministic polynomial(NP) problem and inspired by machine learning,this paper uses genetic algorithm(GA) to model the phase values of the array elements.The equivalent channel matrix is introduced by the product of the channel matrix and the analog precoding matrix.In consideration of the inter-user interference,the digital precoding is carried out with the maximum signal to interference and noise ratio criterion.Simulation results verify that the proposed scheme can approximate the performance of the all-digital precoding matrix.