Doctoral Thesis Multipath Tracking and Prediction for Multiple-Input Multiple-Output Wireless Channels

In this work we develop and study a framework for tracking and prediction of multipath components for wireless MIMO channels. The proposed methodology is a multi-stage procedure that relies on the concept of hypermodels, which capture the dynamics for each multipath. First the individual multipaths are resolved and extracted. In this work we also develop a new estimation algorithm based on the Evidence Procedure and the SAGE algorithm that allows to determine the number of multipath components. The extracted components are then tracked and predicted over time using hypermodels, which are build iteratively, as the tracking proceeds. For prediction we use linear as well as nonlinear hypermodels. We find that linear predictors are more efficient since they are adapted faster. With only 3 coefficients we achieve prediction horizons up to 3 times the wavelength λ for real-world measured data, as compared to 1.5λ reported so far in the literature.