Channel Prediction For Compensating Non-Reciprocity Of Time-varying TDD MIMO Channel

Multiple Inputs Multiple Outputs (MIMO) technique is regarded as a breakthrough in wireless communication. It can increase the system capacity and improve the transmission reliability. In a TDD MIMO system, Channel State Information (CSI) of downlink can be obtained from uplink signal using channel reciprocity, and utilized for design downlink transmission schemes to maximize downlink capacity. However, downlink performance will be degraded due to channel impairments especially in the fast fading scenarios when the CSI obtained from uplink is used for downlink transmission directly. This thesis addresses channel prediction for compensating non-reciprocity of time-varying TDD MIMO channels.The main contents are as follows:1A time-varying TDD MIMO model is established. Then, downlink capacity of SVD-based system is used to evaluate the effect of time-varying channel on channel reciprocity. Simulation results show that system capacities are severely degraded over time-varying channel, and channel reciprocity is destroyed.2The autoregressive (AR) model based and least squares support vector machines (LS-SVM) based channel prediction methods are given to compensate non-reciprocity of time-varying channels. With the same training sequence, LS-SVM based channel prediction has better prediction performance, while AR based channel prediction has less complexity.3A channel prediction method by combing long-range and short-range channel prediction is proposed to compensate non-reciprocity of time-varying channels. The long-range prediction uses frame length as the interval, and the short-range prediction uses sub-frame length. The proposed method had better prediction performance than the short-range prediction. The influence of channel estimate errors on channel prediction has been evaluated. Simulation results show that the proposed method can compensate non-reciprocity of time-varying channel, and avoid error accumulation.