Zero-Forcing Per-Group Precoding for Robust Optimized Downlink Massive MIMO Performance

In this paper, we propose a new, robust near-optimal mutual information combined zero-forcing per-group precoding (ZF-PGP) method for finite alphabets with imperfect channel state information (CSI) at the receiver. ZF-PGP achieves very high gains over ZF precoding techniques, while it simultaneously offers individually separate streams to reach individual user equipment (UE), i.e., it obliterates the need for coordinated, joint decoding by the group's UEs. The proposed robust ZF-PGP scheme comprises an inner linear beamformer (BF), which relies on second-order CSI and is based on the massive MIMO-based joint division and multiplexing for finite alphabets (JSDM-FA) to form orthogonal user groups, and an outer linear precoder that relies on instantaneous group CSI to form a per-group precoding within groups (PGP-WG) with separate data streams to each user. As we show by multiple examples, for a uniform linear array (ULA), the proposed ZF-PGP technique is very efficient, near-optimal design with reduced complexity at the UEs. Due to JSDM-FA second-order CSI inner beamforming, the groups are easily identified. We also demonstrate the robustness of the proposed precoding techniques to channel estimation errors, showing minimal performance loss.