Distributionally Robust Outlier-Aware Receive Beamforming

This paper investigates the robust receive beamforming problem in wireless communication from the perspective of trust-worthy machine learning. The inaccuracies of the signal transmission model, i.e., outliers in the received signals and channel uncertainties, are quantified by the distributional de-viations from the nominal distribution induced by the nominal signal model. The worst-case signal estimation performance (i.e., worst-case mean-squared error) is minimized to achieve robustness against the above two types of distributional in-accuracies. The resultant receive beamformers incorporate clipping operations to mitigate the adverse effects caused by outliers. Experiments validate the effectiveness of the proposed robust receive beamforming method in suppressing outliers and combating channel uncertainties.