毫米波大规模MIMO系统中基于PAST算法的低复杂度混合预编码

在毫米波大规模MIMO系统中,一般采用混合模拟和数字预编码替代全数字预编码来减少射频链和能量消耗.然而,在计算最优无约束混合预编码时,奇异值分解(SVD)具有较高的复杂度.因此,提出了一种基于投影近似子空间跟踪(PAST)的低复杂度混合预编码算法.该算法在计算每个子速率的最优无约束混合预编码时,利用PAST算法估计需要的右奇异矩阵部分主要列向量,从而避免了高复杂度的SVD过程.仿真结果表明,不论是在全连接、混合连接还是在子连接系统结构中,该算法在频谱效率上都接近基于SVD的混合预编码,并且随着发送天线数的增加,提出的算法的复杂度和耗时远低于基于SVD的混合预编码.同时该算法的系统误码率较小,具有较好的可靠性.

Low-complexity hybrid precoding based on PAST for millimeter wave massive MIMO system

Millimeter wave(mmWave) massive MIMO system generally adopts hybrid precoding combining digital and analog precoder as an alternative of full digital precoding to reduce RF chain and energy consumption. However, the singular value decomposition(SVD) could bring high complexity in calculating the optimal unconstrained hybrid precoder. Therefore, this paper proposed a low complexity hybrid precoding algorithm based on projection approximate subspace tracking(PAST). The optimal unconstrained hybrid precoder of each sub-rate, that was, the required major column vectors of the right singular matrix were estimated with the PAST algorithm to avoid the high complexity of SVD process. Simulation results demonstrate that the proposed algorithm approach the performance of the SVD-based hybrid precoding in terms of spectral efficiency for full-connected, hybrid-connected and sub-connected system structure, and as the increase of the number of transmitting antennas, the complexity and time consumption of the proposed algorithm are significantly lower than that of SVD-based hybrid precoding. Meanwhile, the proposed algorithm has smaller bit error rate(BER) and better reliability.