Theoretical performance of a class of space-time adaptive detection and training strategies for airborne radar

First generation airborne radar systems were non-adaptive, performing such operations as moving target indication (MTI), synthetic aperture radar (SAR) imaging, and displaced phased center array (DPCA) data processing. In most cases the processing was separate in space and time (Doppler). Optimal joint space-time adaptive processing (STAP) methods for target detection and parameter estimation have been known for years but were computationally infeasible. Promising hardware technologies, however, have encouraged a revisitation of these optimal methods. The efforts of the DARPA sponsored Mountaintop Program brought to surface some of the weaknesses of these algorithms (which were derived and therefore only optimal under rather ideal assumptions rarely satisfied in the real world). We consider the theoretical performance analysis of a class of STAP detection algorithms under ideal and non-ideal conditions including target steering vector mismatch, sidelobe targets and inhomogeneities, and the impact two of training strategies (i) sliding window with de-emphasis and (ii) power selected training. The detection algorithms considered include the classical adaptive matched filter (AMF), the generalized likelihood ratio test (GLRT), and the more contemporary adaptive cosine estimator (ACE), and the 2-D adaptive sidelobe blanker (ASB).