Low-Complexity Hybrid Linear/Tomlinson-Harashima Precoding for Downlink Large-Scale MU-MIMO Systems

In this paper, we propose a novel low-complexity hybrid linear/Tomlinson-Harashima precoder (HL-THP) for downlink large-scale multiuser multiple-input multiple-output (MU-MIMO) systems. The proposed precoder comprises an inner linear precoder which is designed based on only the second order statistics of the channel state information (CSI) and outer THPs which exploit the instantaneous overall CSI of the cascade of the actual channel and the inner precoder. Thereby, the user terminals (UTs) are divided into groups, where for each group a THP successively mitigates the intra-group interference, whereas the inter-group interference is canceled by the inner linear precoder. Our analytical and simulation results show that the proposed HL-THP achieves a substantially higher average rate per UT and a considerably lower bit error rate (BER) than the linear regularized zero-forcing (RZF) precoder. Moreover, both the average achievable rate per UT and the BER of the proposed HL-THP are close to those of the conventional THP. In addition, we analyze the computational complexity of the proposed HL-THP in terms of the required number of floating-point operations. These results reveal that the proposed HL-THP has a much lower computational complexity than the conventional THP and RZF precoders, respectively. This and its high performance make the HL-THP an excellent candidate for practical systems.