Cooperative multi-cell MIMO downlink precoding for finite-alphabet inputs

This work studies the design of linear precoders for cooperative multi-cell MIMO downlink systems with finite alphabet inputs. Traditionally, multi-cell MIMO downlink precoder designs rely on Gaussian input assumption, which may lead to performance loss when the true inputs admit discrete non-Gaussian symbols. In this work, we present optimized precoders by maximizing weighted sum rate (of finite-alphabet-input) under a set of single base station power constraints. Specifically, we propose a simple gradient algorithm for general multi-cell MIMO downlink channel and a block diagonalization gradient algorithm while supporting interference cancellation.