A Low Complexity Near Maximum Likelihood VBLAST Algorithm for MIMO Systems

In this paper, we propose a new list-based detection algorithm to achieve near maximum likelihood (ML) performance with low computational complexity for multiple-input multiple-output (MIMO) systems. First, we use the minimum mean square error (MMSE) detection to yield an initial bit decision and the corresponding bit log-likelihood ratio (LLR) for a specified layer. Next we construct a list of LLR-based candidate error vectors for the specified layer, and then apply independent reduced-complexity sub-detectors to the remaining layers after cancelling the contribution of each candidate vector of the list. Finally the most likely vector is returned among all visited decisions. Simulation results show that the proposed method can approach identical performance to that of ML detection but only 0.15% complexity for (4,4) system using uncoded 16-QAM.