Performance analysis of low-complexity channel prediction for uplink massive MIMO

Delayed channel state information (CSI) degrades the system performance and predictor can mitigate the effects of outdate CSI. In massive multiple input multiple output (MIMO) systems with large dimensional channel vectors, low-complexity prediction can reduce operation time and process latency. This study adopts a low-complexity channel predictor based on polynomial fitting for the massive MIMO system. Compared with the conventional Wiener predictor, it does not need statistical channel estimation and avoids matrix inversion. The authors derive the approximate signal-to-interference-plus-noise ratio (SINR) with predicted channel information and the approximate gaps of the average rate per user between using perfect CSI, the predicted CSI provided by Wiener predictor and polynomial fitting, respectively, in the uplink massive MIMO system. The authors also analyse the normalised mean square error of prediction. The performance is investigated in a more practical and general angle of departure spectrum model with a concentration direction and a spreading factor. Simulations validate that the SINR approximations are tight, and show that the polynomial fitting with a proper prediction order can achieve a satisfying performance, when the concentration direction and the spreading factor are small.