Multi-objective optimization of reflector antennas using kriging and probability of improvement

This paper presents the multi-objective optimization design of a blocked prime focus reflector system using a Kriging response surface approximation technique with adaptive sampling. Samples are added to the model by selecting those with the highest probability of improving the current estimate of the Pareto front. The problem is especially difficult due to the multi-modal nature of the Pareto front, and a good estimate is achieved using only a modest number of full wave simulations.