Channel Correlation Matrix Based Tomlinson-Harashima Precoding for Downlink Multi-User Transmission

Tomlinson-Harashima precoding (THP) is a kind of nonlinear precoding schemes, which shows impressive transmission performance gain than the linear precoding schemes. However, conventional THP precoding requires accurate channel information at the transmitter to maintain satisfactory performance, such as the multiple input multiple output (MIMO) downlink channel matrix, thus doing nonlinear precoding with precoding granularity of one sub-carrier under frequency selective channel will lead to a limitation of its performance and application. In this paper, for enhancing the performance of conventional THP, we consider to do THP precoding based on channel correlation matrix R for multi-user MIMO systems, which makes better use of the statistical property of the fading channel. We propose a class of novel channel correlation matrix based THP schemes, termed as THP-R and THP-R(E). We also consider the practical case of increasing the precoding granularity under frequency selective channel. Furthermore, we present the analytical analysis of how the proposed THP-R and THP- R(E) schemes are working based on the channel correlation matrix R. Simulation results exhibit that the proposed THP- R and THP-R(E) schemes provide significant performance gain than conventional THP as the precoding granularity increases. THP-R and THP-R(E) show robust performance not only with different precoding granularity, but also with the latency of the channel.