Adaptive differential feedback in time-varying multiuser MIMO channels

In the context of a time-varying multiuser multiple-input-multiple-output (MIMO) system, we design recursive least squares based adaptive predictors and differential quantizers to minimize the sum mean squared error of the overall system. Using the fact that the scalar entries of the left singular matrix of a Gaussian MIMO channel becomes “almost” Gaussian distributed even for a small number of transmit antennas, we perform adaptive differential quantization of the relevant singular matrix entries. Compared to the algorithms in the existing differential feedback literature, our proposed quantizer provides three advantages: first, the controller parameters are flexible enough to adapt themselves to different vehicle speeds; second, the model is backward adaptive i.e., the base station and receiver can agree upon the predictor and variance estimator coefficients without explicit exchange of the parameters; third, it can accurately model the system even when the correlation between two successive channel samples becomes as low as 0.05. Our simulation results show that our proposed method can reduce the required feedback by several kilobits per second for vehicle speeds up to 20 km/h (channel tracker) and 10 km/h (singular vector tracker). The proposed system also outperforms a fixed quantizer, with same feedback overhead, in terms of bit error rate up to 30 km/h.