Robust Multi-Antenna Multi-User Precoding Based on Generalized Multi-Unitary Decomposition With Partial CSI Feedback

In this paper, we present a novel matrix decomposition method, called generalized multi-unitary decomposition (GMUD), and use it to achieve robust multi-antenna multi-user multiple-input-multiple-output (MIMO) precoding with limited and partial channel information feedback. GMUD transforms complex matrix {H} into H = P_θ, _rR_rQ_θ,r^H, where R_r is a lower triangular matrix, and P_{θ, r} and Q_{θ, r} are unitary matrices. A unique feature of GMUD is that it gives multiple solutions of the unitary matrices P_{θ, r} and Q_{θ, r} based on two parameters, namely the direction parameter θ and magnitude parameter r. This property enables the GMUD precoder to steer the precoding vectors of different users to jointly optimize their performance. Compared with regularized-inverse precoding (an existing multi-user MIMO precoding technique that we have extended to handle multiple receive antennas), the advantages of GMUD precoding are that it requires only partial channel state information (CSI) from the users, and its performance is robust to CSI quantization.