Tomlinson-Harashima Precoding for Multiuser MIMO Downlink with Imperfect Channel State Information

Abstract

Most contributions on nonlinear Tomlinson-Harashima precoding (THP) assume complete channel state information (CSI) at the transmitter. However, complete CSI is not available in many wireless systems, and the conventional THP is sensitive to channel mismatch. This article investigates the THP design with imperfect CSI at base station (BS) in spatially correlated multiuser multiple-input multiple-output (MIMO) downlink. By exploiting Bayes' theorem and the channel statistics, a robust minimum mean-square error (MMSE) THP is proposed, which can effectively decrease the unfavorable influence caused by spatial correlation and channel estimation errors. Simulation results are used to show the performance advantage attained by the proposed THP.