Non-linear vector-perturbation precoding for multi-user downlink under quantized CSI

This paper focuses on the design of vector perturbation (VP) precoding for multi-user multiple-input singleoutput downlink transmission under quantized channel state information. Each receiver decomposes its downlink channel vector in forms of channel direction information (CDI) and channel magnitude information (CMI) for feedback to the transmitter. Under quantized CDI and quantization error statistics, closed-form expressions to the mean-squared-error (MSE) between channel input and output when (i) perfect CMI available to the transmitter and (ii) only CMI statistics known at the transmitter, are derived. We then propose a unified approach to design the MSE minimization based VP precoders. Bit error rate simulation results indicate that the proposed VP precoder designs are less sensitive to quantization errors and CMI availability helps to improve the performance.