Exploiting Spatio-Temporal Correlations in MIMO Wireless Channel Prediction

We investigate prediction algorithms that exploit both temporal and spatial correlations in MIMO correlated narrowband fading wireless channels. We first derive the optimal two dimensional minimum mean square error (2D-MMSE) prediction filter that maximally exploits the available temporal and spatial correlations. We then propose a lower complexity 2-step prediction algorithm, which first exploits temporal correlations using a classical single-input single-output (SISO) MMSE time-domain prediction filter for each entry in the MIMO channel, followed by an MMSE spatial smoothing step to exploit the spatial correlations. Compared to the SISO and specular prediction approaches, this approach either achieves lower MSE with a slight increase in complexity, or comparable MSE with lower complexity, in a wide range of wireless channel conditions. The same advantage holds for the average mutual information.