Vector quantized CSI prediction in linear multi-user MIMO systems

In this paper, we discuss the methods of improving sum-rates of linear multi-user multiple-input multiple-output (MIMO) systems employing vector quantization (VQ) algorithms for encoding channel state information (CSI). We consider an approach where the mobile receiver estimates the downlink channel and chooses one of the predefined channel characterization CSI codewords whose indices are transmitted back to the base station transmitter. The algorithms described here are aimed at combating bit errors, erasures and delays occurring in CSI feedback channels. The proposed solution is to design transmitter-side CSI index prediction algorithms that allow the system to alleviate feedback channel imperfections. Moreover, the proposed techniques can be used to lower the required feedback rate, while keeping the system sum-rate at the required level. We demonstrate the performance of our algorithms in broadcast MIMO systems with and without CSI prediction. The simulation results show that significant sum-rate gains are achievable for a large range of channel Doppler rates and feedback delays.