Improved Gaussian Process Modelling of On-Axis and Off-Axis Monostatic RCS Magnitude Responses of Shoulder-Launched Missiles

An improved Gaussian-process-based technique is described for modelling both on-axis and off-axis monostatic RCS magnitude responses of shoulder-launched missiles. The RCS responses are complicated, oscillatory, quasi-periodic functions of frequency, with the oscillation periods being related to the spacings of the scatterers comprising the missiles. The updated modelling approach employs a spectralmixture covariance function, whose components explicitly include distinct oscillations. The proposed technique is evaluated by means of two example missiles. For the six rotations considered, average predictive normalised RMSE ranged from 0.34% to 0.87% and from 0.95% to 1.53% for the two missiles respectively.