Signature Prediction and Uncertainty Analysis for Recognition Applications

Use of full wave solvers such as TEMPUS is still prohibitively expensive for generating Automatic Target Recognition (ATR) data domes for full scale targets at X-band frequencies. HyPerComp is currently developing a hybrid procedure based on the “CrossFlux” approach that combines the fast high frequency Xpatch with full wave TEMPUS to significantly reduce the computational burden. In addition, HyPerComp is developing “uncertainty” measures for computational simulation to feed into the ATR data base, and is also developing a framework to implement an innovative concept described as Reduced-Order Basis Method (RBM) that can significantly reduce the computational resource requirements by two to four orders of magnitude from its current levels. The main simulation tools are TEMPUS and Xpatch. The enabling technologies are RBM, hybrid CrossFlux, uncertainty models, and potential algorithmic improvements to high order full wave procedures and expected increase in hardware power. If all of these come together, within the next 3-5 years, HyPerComp expects a revolutionary reduction in computational resource requirements. Under this proposed effort, HyPerComp plans to work with Professor Hesthaven of Brown University in the development, implementation and validation of the reduced-order basis method, and continue to work with SAIC-DEMACO in further validation and demonstration of “CrossFlux” hybrid procedures.