Abstract:
An efficient technique for modeling radar cross section magnitude responses versus frequency is presented. The technique is based on Gaussian process regression and makes it possible to significantly reduce the number of expensive computer simulations required to accurately resolve these responses. Examples of two missiles are used to evaluate the proposed technique. Average predictive normalized root-mean-square errors (RMSEs) of 1.24% and 1.63% were obtained, with the worst RMSE being less than 2.2%. These results were significantly better than results obtained with alternative techniques, including geometric theory of diffraction-based modeling and support vector regression.
