Equal-rate QR decomposition based on MMSE technique for multi-user MIMO precoding

In recent years, the research on multiple-input multiple-output (MIMO) wireless communications has attracted much interest. This study investigates precoding techniques for multi-user MIMO communications. By applying QR decomposition to augmented channel matrix based on minimum mean squared error (MMSE) approach plus Tomlinson-Harashima precoding (THP) and equal-rate power allocation, we show that the proposed equal-rate QR-MMSE-THP precoder has better performance than the one based on zero-forcing (ZF) channel matrix, called equal-rate QR-ZF-THP precoder, and the conventional MMSE precoder. In addition, sorting strategies can be adopted to enhance performance with more freedom for QR decomposition compared to block diagonalized decompositions such as block diagonalized-geometric mean decomposition (BD-GMD) and BD-GMD-MMSE schemes while complexity can still be kept practical. Also, the sorting strategies such as full sorting (FS), per-layer sorting (PLS) and per-user sorting (PUS) are discussed and compared in this paper. Simulation results show that for large MIMO systems, the proposed equal-rate QR-MMSE-THP with PLS strategy outperforms the equal-rate BD-GMD-MMSE-THP precoder with PUS.