Multiuser Precoding With Limited Cooperation for Large-Scale MIMO Multicell Downlink

Abstract

Multiple base stations (BSs) equipped with large antenna arrays can perform cooperative downlink transmissions using beamforming for cell-edge users to provide high data rate services. However, due to an excessive overhead for the channel state information (CSI) exchange between BSs, the cooperation with full CSI may not be practical for large antenna arrays. Thus, based on an existing precoding approach, two different precoding approaches with partial cooperation, where no CSI exchange is required, are proposed. One of them (called Type III precoding) has a better performance, and its performance can be further improved by user allocation, which becomes possible by deriving a closed-form expression for transmission power. For low-complexity implementations, an algorithm is derived for the user allocation based on the greedy algorithm. Under certain channel conditions, closed-form expressions for average transmission powers are also derived to see asymptotic performance. Based on them, we can show that the average transmission power becomes $Q$ times higher than that with full cooperation if precoding is carried out with local CSI only, where $Q$ is the number of BSs in cooperation, which would be considered as a penalty due to partial cooperation. Through simulations, it is shown that this penalty (i.e., the increase of transmission power) can be reduced by user allocation.