Prediction for time-varying SVD systems

The performance of a system using singular value decomposition (SVD) over a multiple-input multiple-output (MIMO) channel is dependent on the accuracy of the channel state information (CSI) at both the transmitter and receiver. In time division duplex (TDD) systems the reciprocal nature of a channel means that the CSI obtained during the uplink can be applied to data transmission on the downlink. However, for a time-varying channel the estimated CSI at the transmitter is incorrect due to an inherent delay between the estimation of the CSI and the transmission of data. This severely degrades typical SVD system performance. This paper proposes to reduce this degradation through prediction of the CSI. Prediction can also be applied to the preceding matrix. However, when using a finite impulse response (FIR) filter the best performance occurs when prediction is applied to the CSI followed by the SVD.