Date Added: Dec 2009
In this paper, the authors develop a family of approximate Maximum-Likelihood (ML) detectors for multiple-input multiple-output systems by relaxing the ML detection problem using constellation-specific polynomial constraints. The resulting constrained optimization problem is solved using a penalty function approach. Moreover, to escape from the local minima, which improves the detection performance, a differential equation algorithm using classical mechanics is proposed. Simulation results show that the polynomial constrained detector performs better than Least-Squares (LS) detector.