This paper attempts to jointly optimize the hybrid precoding (HP) and intelligent reflecting surfaces (IRS) beamforming matrices in a multi-IRS-aided mmWave communication network, utilizing the Alamouti scheme at the base station (BS). Considering the overall signal-to-noise ratio (SNR) as the objective function, the underlying problem is cast as an optimization problem, which is shown to be non-convex in general. To tackle the problem, noting that the unknown matrices contribute multiplicatively to the objective function, they are reformulated into two new matrices with rank constraints. Then, using the so-called inner approximation (IA) technique in conjunction with majorization-minimization (MM) approaches, these new matrices are solved iteratively. From one of these matrices, the IRS beamforming matrices can be effectively extracted. Meanwhile, HP precoding matrices can be solved separately through a new optimization problem aimed at minimizing the Euclidean distance between the fully digital (FD) precoder and HP analog/digital precoders. This is achieved through the use of a modified block coordinate descent (MBCD) algorithm. Simulation results demonstrate that the proposed algorithm outperforms various benchmark schemes in terms of achieving a higher achievable rate.

翻译：本文尝试在基站采用Alamouti方案的多智能反射面辅助毫米波通信网络中，联合优化混合预编码与智能反射面波束赋形矩阵。以总体信噪比为目标函数，该问题被构建为一个优化问题，并证明其通常具有非凸性。为解决该问题，注意到未知矩阵以乘积形式影响目标函数，将其重构为两个具有秩约束的新矩阵。随后，结合内逼近技术与最大化-最小化方法，对这些新矩阵进行迭代求解。从其中一个矩阵可有效提取智能反射面波束赋形矩阵。同时，混合预编码矩阵可通过旨在最小化全数字预编码器与混合模拟/数字预编码器间欧氏距离的新优化问题独立求解，该过程通过改进的块坐标下降算法实现。仿真结果表明，所提算法在实现更高可达速率方面优于多种基准方案。