Dynamic Nulling-and-Canceling for Efficient Near-ML Decoding of MIMO Systems

摘要

It is known that conventional nulling-and-canceling (NC) detection for multiple-input/multiple-output (MIMO) systems cannot exploit all of the available diversity, and, thus, its performance is significantly inferior to that of maximum likelihood (ML) detection. Conventional NC employs the layerwise postequalization signal-to-noise ratios (SNRs) as reliability measures for layer sorting. These SNRs are average quantities that do not depend on the received vector. In this paper, we propose the novel dynamic nulling-and-canceling (DNC) technique that uses approximate a posteriori probabilities as measures of layer reliability. The DNC technique is a minimum mean-square error (MMSE) nulling scheme combined with an improved "dynamic" layer sorting rule that exploits the information contained in the current received vector. We calculate the error probability of DNC for a simple special case and show that it is upper bounded by the error probability of conventional NC. Simulation results are presented for spatial multiplexing systems and for systems using linear dispersion codes. It is demonstrated that the DNC technique can yield near-ML performance for a wide range of system sizes and channel SNRs at a fraction of the computational complexity of the sphere-decoding algorithm for ML detection