Accurate channel state information (CSI) is essential at the base station (BS) in order to serve multiple users simultaneously on the same time-frequency resources. Explicit low-rate CSI feedback from each user terminal to the BS is indispensable in frequency-division duplexing (FDD) multipleinput multiple-output (MIMO) hybrid precoding systems. In the presence of correlated Rician fading channels with arbitrary rank, and in order to reduce the feedback overhead, a unified adaptive statistical codebook is designed by leveraging the long-term first and second order channel statistics. Moreover, as the feedback overhead scales linearly with the number of BS antennas, which introduces a bottleneck in FDD massive MIMO systems, an adaptive two-stage reduced dimensionality codebook design is developed. The main idea stems from the fact that the high dimensional analog beamforming stage could get away with coarse CSI, e.g., statistical CSI, while the low dimensional baseband precoding stage acquires finite-rate quantized CSI of the reduced dimensionality channel to eliminate multiuser interference. These steps reduce significantly the CSI training and feedback overhead to scale linearly with the number of radio-frequency chains instead. For comparison purposes, the two extremes of full and statistical CSI at the BS are also included. The performance gap between these two extremes justifies the importance of the proposed two-stage reduced dimensionality adaptive statistical codebook.
|Name||ICC: IEEE International Conference on Communications (ICC): Proceedings|