Reducing the effects of training data heterogeneity in multistatic MIMO radar

A MIMO Multistatic radar system consists of multiple bistatic pairs working in potentially different configurations. Due to the relative motion between platforms, the clutter traces are, in general, non-overlapping, and the spectral centers are dispersed in the angle-Doppler domain. This makes the training samples non-representative which adversely affects the system performance. In this paper, we explore existing techniques to compensate for training data heterogeneity, and characterize the performance of each technique based on the metrics of SINR loss and compare the performance of the Parametric Adaptive Matched Filter (PAMF) with diagonally-loaded Sample Matrix Inverse (DL-SMI) subspace detector. We also show that PAMF is resilient to training data heterogeneity and provides acceptable performance compared to DL-SMI when no compensation is applied to the training data.