Deep Learning for Multi-User MIMO Systems: Joint Design of Pilot, Limited Feedback, and Precoding

In conventional multi-user multiple-input multiple-output (MU-MIMO) systems with frequency division duplexing (FDD), channel acquisition and precoder optimization processes have been designed separately although they are highly coupled. This paper studies an end-to-end design of downlink MU-MIMO which include pilot sequences, limited feedback, precoding. To address this problem, we propose a novel deep learning (DL) framework jointly optimizes the feedback information generation at users base station (BS). Each procedure in is replaced by intelligently multiple neural networks (DNN) units. At BS, network generates sequences helps obtain accurate state information. each user, operation carried out distributed manner individual user DNN. Then, another BS DNN collects from determines MIMO precoding matrices. A joint training algorithm proposed to optimize all units manner. addition, strategy can avoid retraining for different sizes scalable proposed. Numerical results demonstrate effectiveness DL compared classical techniques other schemes.