Manifold Optimization Approach for Data Detection in Massive Multiuser MIMO Systems

Abstract

When the number of base station (BS) antennas is considerably larger than the number of user terminals (UTs), a simple linear minimum mean-square-error (LMMSE) data detection algorithm is near-optimal for uplink massive multiuser multiple-input and multiple-output (MIMO) systems. However, the LMMSE detector suffers significant performance loss when the number of UTs is comparable to the number of BS antennas. To alleviate the performance loss caused by the decrease in the BS-antenna-to-UT ratio, known data detection algorithm, namely “TASER,” can yield bit error rate (BER) performance close to that of the optimal detector in symmetric massive multiuser MIMO systems, where the BS-antenna-to-UT ratio is one. However, this TASER algorithm can only handle binary and quadratic phase-shift keying (PSK) modulations. Moreover, the computational complexity of the TASER algorithm is still remarkably high. Thus, a novel data detection algorithm is proposed in this study to reduce the computational complexity of the detection algorithm while achieving acceptable BER performance. The proposed algorithm is developed from the retraction-based Riemannian manifold optimization (RMO) framework for symmetric massive multiuser MIMO systems. Unlike the TASER-based detector, the proposed RMO-based detector can be applied for general $M$ -ary PSK modulations. Simulation results show that the proposed detector significantly outperforms the LMMSE detector in terms of BER performance. In addition, the required complexity of RMO is substantially lower than that of TASER. These findings indicate that the proposed RMO-based detector exhibits a highly desirable tradeoff between BER performance and complexity.