Performance of low complexity receivers for massive MIMO with channel estimation and correlation

In an uplink massive multiple-input-multiple-output (MIMO) deployment with distributed single-antenna users and a large base-station array, we investigate the performance of low complexity Vertical Bell Laboratories Layered Space Time (C-V-BLAST) and zero forcing (ZF) receivers. Unlike previous work which assumes that the channel is perfectly known, we focus on the realistic scenario where the MIMO channel is actually estimated. We consider two channel estimation strategies, pilot-based maximum likelihood (ML) which has a closed-form and semi-blind (SB) estimation which is performed iteratively using steepest-descent. To the best of our knowledge, the performance gains of SB estimation over ML estimation, and the resulting complexity tradeoffs have not been investigated in previous works in the context of massive MIMO with low complexity receivers. The motivation of our study is to fill this gap. Our work shows that the SB estimation strategy yields a significant improvement in symbol-error-rate (SER) performance, with ZF outperforming the C-V-BLAST. We also investigate the impact of channel correlation on system performance.