MULTIUSER MIMO SYSTEMS EMPLOYING SEMIDEFINITE RELAXATION OF MAXIMUM LIKELIHOOD DETECTION

It is well known that the optimal algorithm for detection of multiuser systems is the maximum likelihood ML sequence detector. However, the ML detector needs to solve a hard combinatorial optimization problem with complexity that increases exponentially with respect to the number of active users. For this reason, in the last decade semidefinite programming SDP has been used to relax the exponentially complex ML detection problem into an SDP problem that features polynomial worst-case complexity. This relaxation was considered either for multiuser single-input single-output MU-SISO systems or for single user multi-input multi-output SU-MIMO systems. This paper investigates the performance of the semidefinite relaxation SDR detector in multiuser multi-input multi-output MU-MIMO systems. Simulation results show that the SDR detector outperforms several existing suboptimum detectors, in terms of bit error rate BER, even in the existence of the near-far problem. Furthermore, the SDR detector exhibits slight degradation relative to the optimum ML detector.