Adaptive Search in QRD-M for Complexity Reduction in MIMO X Channel

3GPP LTE introduces multiple-input multiple-output (MIMO) spatial multiplexing to increase the peak spectrum efficiency. But the effectiveness of the cell edge users is degraded due to inter-cell interference (ICI). Thus, 3GPP long term evolution-advanced (LTE-A) further applies coordinated multi-point (CoMP) to solve this problem. In multi-Tx multi-Rx (MTMR) MIMO system, each eNodeB transmits data to different users simultaneously. When applying multi-user MIMO precoding to eliminate the intra-cell interference, the data detection scheme of MTMR MIMO is the same as traditional single-eNodeB single-user MIMO. The optimal detection of MIMO system is understood to be achieved by maximum-likelihood (ML) detection in theory. However, the computation complexity of ML detection is prohibitive. In this paper, we use minimum-mean-square-error based ordered successive interference cancellation (MMSE-OSIC) solution as an initial guess. Then, we use the SNR and SINR under which these initial estimations are derived as a reference to weight the effectiveness of these solutions. This effectiveness is used to decide the search range of the consequent ML estimation. In the ML estimation, QR decomposition is used to diagonalize the channel matrix. In each level, only limited points around the MMSE-OSIC solution point in the M-ary QAM constellation are examined as compared with a full search. At the end of each level, only M most significant symbol combinations are retained.