Hybrid Preconditioned CG Detection with Sequential Update for Massive MIMO Systems

Abstract

Massive Multi-Input Multi-Output (MIMO) is one of the key technologies for the fifth generation communication systems. Conjugate Gradient (CG) algorithm approximates the minimum mean-square error (MMSE) in an iterative manner, which avoids full matrix inversion. Pre-conditioned CG (PCG) was presented to improve the robustness of CG method. However, for the PCG, a sparse matrix inversion is still required in preprocessing and the performance is only comparable to MMSE. In this paper, a hybrid PCG algorithm (HPCG) with sequential update is proposed with superior performance and low complexity. The preconditioned matrix is replaced by a diagonal matrix by exploring its characteristics, which avoids matrix inversion and incomplete Cholesky factorization. Besides, to improve the bit error performance, a sequential update strategy is employed for estimated signals after PCG detection. For a MIMO system with 128 receive antennas, simulation results show the proposed HPCG algorithm outperforms MMSE by 0.25 dB to 1.5 dB under different numbers of users. Based on the channel hardening theories, several signal vectors can be transmitted in the same channel condition. When 10 signal vectors are considered, compared to the other CG based algorithms, the overall complexity of HPCG can be reduced by 3.9% to 56%.