Iterative Time-Varying Channel Prediction Based on the Vector Prony Method

This paper proposes channel prediction algorithms, based on the vector Prony method, tackling the channel state information (CSI) acquisition problem in fast time-varying uplink and downlink channels of the time division duplex (TDD) in massive multiple-input, multiple-output (MIMO) systems. For the uplink system, we show that the CSI obeys the Prony equation in theory, and we aim to overcome the instability of solving the Prony equation via iterating the Prony coefficients based on CSI re-estimated from signals received during data transmission. For the downlink system, a method to generate precoders is constructed based on the prediction of downlink CSI. Experiments are performed according to the fast fading full-dimensional multiple-input multiple-output (FD-MIMO) channel model of 3GPP Standards. The results demonstrate that performance is improved remarkably in uplink transmissions and approaches the theoretical optimal scheme in downlink transmissions.