Low-complexity Kaczmarz precoding in DL massive MIMO with partial CSI and correlation

Abstract

Abstract The capacity achieved by three low-complexity precoders is compared considering a single-cell massive MIMO (M-MIMO) broadcast channel; these linear precoders include the conventional zero-forcing (ZF) beamforming, regularized channel inversion (RCI) precoding, and a precoding version based on the iterative randomized Kaczmarz algorithm (rKA). The capacity vs complexity tradeoff is analyzed in the downlink (DL) with a base-station (BS) deploying massive uniform linear array (ULA) antennas, while mobile terminals (MT) are equipped with single-antenna. Based on the precoding matrix associated with each scheme, we first derive the interference matrix and, consequently, the capacity achieved by each analyzed precoder. Numerical results demonstrate that the rKA’s performance-complexity tradeoff is superior compared to conventional ZF and RCI precoding when operating in M-MIMO scenario, since it holds a relative robustness against system loading increasing, appears to be equally efficient under a simple equal power assignment (EPA) policy and implies in a considerably reduced number of complex operations.