基于黎曼拟牛顿的毫米波MIMO混合预编码算法

Millimeter-wave MIMO hybrid Precoding Algorithm Based on Riemannian Quasi-Newton

针对毫米波MIMO系统混合预编码算法的非凸单位模约束、矩阵耦合问题,文章提出了一种基于黎曼拟牛顿的交替最小化混合预编码(RLBFGS-AltMin)算法。该算法首先引入交替最小化框架(AltMin),有效解决了混合预编码矩阵联合优化困难的问题;其次,利用黎曼流形将模拟预编码部分中的非凸单位模约束转化为搜索空间无约束优化求解;最后,通过改进的梯度计算方式和黎曼梯度下降理论求解出最优数字预编码和模拟预编码矩阵。仿真实验结果表明,所提算法可以获得更接近全数字预编码的频谱效率。

Aiming at the problem that the hybrid precoding algorithm of millimeter-wave MIMO system is constrained by non-convex unit modulus and matrix coupling,this paper proposes an alternating minimization hybrid precoding(RLBFGS-AltMin)algorithm based on Riemannian quasi-Newton.This algorithm first introduces AltMin,which effectively solves the problem of joint optimization of hybrid precoding matrix.Secondly,by using Riemannian manifold,the nonconvex unit modulus constraints in the analog precoding part are transformed into unconstrained optimization in search space.Finally,the optimal digital precoding and analog precoding matrix are solved by the improved gradient calculation method and Riemann gradient descent theory.Simulation results show that the proposed algorithm can achieve spectral efficiency closer to that of full digital precoding.