Selection Based List Detection With Approximate Matrix Inversion for Large-Scale MIMO Systems

Abstract

Large-scale multiple-input multiple-output (LS-MIMO) technology constitutes a foundation for next generation wireless communication systems. Detection techniques are a key issue for practical applications of LS-MIMO. Selection-based list detection is an attractive approach for LS-MIMO systems because of its massively parallelizable nature. In this paper, we propose an improved selection-based list detection algorithm that exploits the channel hardening phenomenon, making it suitable for LS-MIMO. We start by introducing a low latency approximate inversion technique for large dimensional complex matrices, which can be used not only in selection-based list detection but also in many other LS-MIMO detection algorithms. The proposed matrix inversion technique is integrated into linear as well as selection-based list detection algorithms for lower latency and deeper parallelism. Then our analysis of the impact of channel hardening on selection-based list detection motivates the use of an improved ordering scheme for the successive interference cancellation sub-detector. Finally, we compare our improved selection-based list detector with other two state-of-art low complexity LS-MIMO detection algorithms, namely, multistage likelihood ascent search (LAS) and message passing detection (MPD). Computer simulations show that the proposed selection-based list detector performs better than multistage LAS and has just a small fraction of dB performance loss compared with MPD. Because of its good performance and parallelizable nature, the proposed algorithm offers an attractive alternative for detection in practical LS-MIMO systems.