Regularized zero-forcing precoding with non-homogeneous user conditions

This paper is concerned with linear precoding designs in a multiuser downlink system. We consider a multiple-input single-output system with multiple single-antenna user-equipments (UE) experiencing non-homogeneous user conditions, including the channel strength and the background noise power. Assuming perfect knowledge of channel state information and noise power at the base-station (eNB), we propose a new regularized zero-forcing (RZF) precoder, which takes advantage of the non-homogeneous user conditions. Given in a closed-form solution, the proposed RZF precoder outperforms other well-known linear precoders, while achieving a close performance to the locally optimal iterative weighted minimization of mean-squared error precoder, in terms of the achievable network sum-rate. We then propose a greedy user selection algorithm in conjunction with the proposed RZF precoder when the number of UEs exceeds the number transmit antennas at the eNB.