Spatial Correlation Based CSI Feedback Reduction in Dual-Polarized Massive MIMO System

Abstract

In this paper, a joint antenna selection and two- stage precoding scheme is proposed to reduce the channel state information (CSI) feedback in dual- polarized massive multiple-input multiple-output (MIMO) system. The proposed scheme can exploit the polarization resource and the spatial correlation from both sides of the transceiver to acquire more accurate CSI with low feedback. First, an antenna selection algorithm is proposed in the pilot training stage to select the least spatially correlated BS antennas, and the acquired partial CSI can be averaged for the CSI of remaining antennas. Both the dimension of the effective uplink channel and the number of necessary radio frequency chains can be reduced. Then, by grouping the spatially correlated users, the two-stage precoding algorithm further divides each CSI coefficient in the effective uplink channel into two categories. Long- term CSI can be reported less frequently while short-term CSI feedback can also be reduced in half due to the isolation between orthogonal polarized directions. The simulation results show that the proposed scheme can enhance both the achievable rates and the energy efficiency of the system.