Precoding for Multicell Massive MIMO Systems with Compressive Rank-q Channel Approximation

In non-cooperative multicell massive (or very large) multiple-input multiple-output (MI MO) systems, the pilot contamination effect in the uplink channel training and the intercell interference in the downlink severely degrade the overall network achievable rates. In this paper, we present a framework to mitigate those effects. Specifically, we propose in the uplink training rank-q channel approximation method based on compressive sensing (CS) to estimate the most dominant singular subspaces of the global multicell MIMO channel matrix with a modest training length. Then, the estimate of the global channel inlbrmation is used to design an intercell-interference-aware (IA) zero-lbrcing (ZF) multicell precoding method in the downlink to mitigate not only the intracell interference but also the intercell interference of the channel. The results obtained using the proposed scheme show a significant improvement in the achievable rates lbr all users in the cells, particularly the cell-edge users, as compared to the existing channel estimation and precoding methods.