Reduced-Cost Constrained Miniaturization of Wideband Antennas Using Improved Trust-Region Gradient Search With Repair Step

Abstract

In this letter, an improved algorithm for electromagnetic (EM)-driven size reduction of wideband antennas is proposed. Our methodology utilizes variable-fidelity EM simulation models, auxiliary polynomial regression surrogates, as well as multipoint response correction. The constraint handling is implicit, using penalty functions. The core optimization algorithm is a trust-region gradient search with a repair step added in order to facilitate exploration of the feasible region boundary. The latter results in considerable reduction of the computational cost of the optimization process. The proposed methodology is demonstrated using two wideband antennas and compared to both direct optimization of the high-fidelity model and the surrogate-assisted framework without the aforementioned repair step. The results indicate superiority of the presented approach. Experimental validation of the fabricated prototypes of the optimized antennas is also provided.