Trellis Precoding for the Multi-User Environment

When transmitting over the multi-user MIMO broadcast channel, the transmitter has both knowledge of, and control over, the interference between users. Methods based on interference cancellation focus only on the transmitter´s knowledge of the interference. This paper proposes a method which takes advantage of the transmitter´s control over the interference among users. This is done by considering all users simultaneously. First, redundancy is introduced into the transmitted data similarly to the trellis shaping method, then the mean square error (MSE) is minimized by exploiting this redundancy. The proposed method is unique in that it considers all users simultaneously, and that the MSE is minimized rather than the transmit power. This method can achieve a BER of 10^-5, only 1 dB above the minimum signal to noise ratio predicted by the capacity of the channel. Depending on the channel parameters, this demonstrates a 1-3 dB improvement over the best reported results. The improvement in the performance is achieved at the price of an increase in the complexity (equivalent to the Viterbi decoding of a trellis with 2^2N states for N transmit antennas).