This paper proposes a computationally efficient algorithm for downlink (DL) multiuser (MU) MIMO systems using a new zero-forcing (ZF) Tomlinson-Harashima precoding (THP) scheme maximizing a sum-rate. The ZF-THP is more computationally efficient than minimum mean square error (MMSE) THP; however, its bit error rate performance is inferior to the one. User stream permutations, which determine the order of the interference cancellation at the transmitter, is incorporated into the ZF-THP; then, we can select a permutation pattern so as to maximize the sum-rate. Since searching for an optimal permutation pattern is computationally cumbersome, we propose an efficient algorithm, which produces almost optimal user stream permutation, for the ZF-THP. We also propose to allocate unequal number of bits to each user to improve DL performance of the MU-MIMO system. Numerical results are provided to show that the proposed ZF-THP scheme outperforms the conventional MMSE-THP scheme.
