Research On Quantized Channel Prediction Methods For TDD-MIMO Systems

In the TDD-MIMO systems, with the help of channel reciprocity, the base stationcan preprocess data according to the channel state information (CSI) obtained by uplinkestimation. It can achieve significant gain in system capacity. However, time-varyingchannel leads to the loss of channel reciprocity, which will degraded the downlinkperformance when the CSI obtained from uplink is used for downlink transmissiondirectly. This thesis addresses methods for compensating non-reciprocity intime-varying TDD-MIMO systems. The main contributions are as follows:1System capacity of TDD-MIMO system is derived using Singular ValueDecomposition (SVD). Simulation results show that time-varying channel destroys thechannel reciprocity and degrades the system capacity of TDD-MIMO system.2The Grassmannian Predictive Coding (GPC) algorithm is proposed tocompensate the channel non-reciprocity in TDD-MIMO systems. Quantization andprediction on the Grassmannian manifold, the GPC algorithm contributes to channelreciprocity compensating effectively and system capacity improving significantly.3The LS-SVM multi-class classifier based prediction algorithm is developed tocompensate the channel non-reciprocity. Simulation results show that by predicting theindex of the CSI quantization in Grassmannian codebook, the algorithm has theadvantage in fast fading scenarios.