毫米波大规模MIMO系统中基于GMD的低复杂度混合预编码

GMD-based low complexity hybrid precoding in millimeter-wave massive MIMO systems

针对毫米波大规模多输入多输出(multiple input multiple output, MIMO)系统中,基于几何均值分解(geometric mean decomposition, GMD)的混合预编码方案计算复杂度高的难题,提出一种基于GMD的低复杂度混合预编码方案。根据交替最小化原则,通过基于阶递归最小二乘的广义正交匹配追踪算法设计具有非凸约束的模拟预编码器;当模拟预编码器确定后,基于最小二乘准则设计数字预编码器。通过在每次迭代中选择多个向量减少最大迭代次数以及利用阶递归最小二乘避免模拟预编码器设计中的矩阵求逆来降低方案复杂度。通过对复杂度及频谱效率和误码率性能的仿真结果分析表明,所提方案在显著降低计算复杂度并提高预编码效率的情况下,性能接近现有基于GMD的混合预编码方案,并且误码率优于传统基于奇异值分解(singular value decomposition, SVD)的混合预编码方案。

Aiming at the high computational complexity of GMD-based hybrid precoding scheme in millimeter-wave massive MIMO systems,we propese a low complexity GMD-based hybrid precoding scheme.According to the principle of alternating minimization,a generalized orthogonal matching pursuit algorithm based on order recursive least squares is firstly used to design an analog precoder with non-convex constraints.Second,when the analog precoder is determined,the digital precoder is designed based on the least squares criterion.The complexity of the scheme is reduced by selecting multiple vectors in each iteration to reduce the maximum number of iterations and using order recursive least squares to avoid matrix inversion in the analog precoder design.Through the analysis of complexity and the simulation results of spectral efficiency and bit error rate performance,this scheme is very close to the existing GMD-based hybrid precoding scheme in the case of significantly reducing computational complexity and improving precoding efficiency,and the bit error rate is better than the traditional SVD-based hybrid precoding scheme.