Research On MIMO - OFDM Channel Prediction Algorithm

Channel prediction is an appealing technique to mitigate the performance degradation due to the inevitable feedback delay of the channel state information (CSI) in modern wireless systems. In the presence of correlation among the transmit or receive antennas, it will improve the performance of channel prediction if utilizing the correlation information. That is, in multiple input multiple output orthogonal frequency division multiplexing (MIMO-OFDM) systems, both the frequency and spatial correlation are beneficial to prediction. Although several studies have been reported on the channel prediction which either considers spatial correlation or frequency correlation, no results have been reported on the MIMO-OFDM systems which take the spatial, frequency, time correlation into account simultaneously to the best of our knowledge.In this dissertation, we aim to propose the MIMO-OFDM channel prediction algorithms which exploit the spatial correlation to improve the prediction performance. There are three main contributions in this dissertation.First, we proposed a frequency-domain MIMO-OFDM channel prediction algorithm. In this part, we propose a trade-off way which exploits the time, space, frequency correlation effectively while maintains a low complexity. First, it is analyzed that the channel correlation can be decoupled into time, space, frequency part. Then based on the separation property of the channel correlation function, three separate one dimension (1-D) filters are used which consider time, frequency, spatial correlation respectively.Second, we proposed a time-domain MIMO-OFDM channel prediction algorithm. In this part, we first propose a MIMO-OFDM channel prediction model in time domain, where the muti-carrier MIMO prediction problem is transformed to the single-carrier MIMO problem by FFT/IFFT. Then we derive two predictors based on the proposed model, which considers and exploits the spatial correlation. Both of the two predictors select data for auto-regressive (AR) modeling in different ways. The first predictor, called forward-stepwise subset (FSS) predictor, chooses the desired data set via minimizing the mean square error (MSE) of prediction model. Yet the second predictor, called reduced-complexity FSS (RCFSS) predictor, chooses the data in a heuristic way, which aims to reduce the computational complexity. Both of the two predictors can exploit the temporal and spatial correlation adaptively.Finally, to further investigate the application of the proposed prediction methods, our algorithms are applied to improve the precoding performance in multi-user MIMO-OFDM systems. When the channel changes rapidly, the outdated error results in significant performance degradation, where the channel prediction can overcome the outdated problem effectively.