Rapid optimization of dielectric resonator antennas using surrogate models

Adjustment of geometry parameters and performance optimization of dielectric resonator antennas (DRAs) is a challenging task that requires electromagnetic (EM) simulations. In particular, DRA design automation is problematic due to high computational cost of accurate, high-fidelity EM simulations; specifically, embedding the EM solver directly into the optimization algorithm (e.g., gradient-based) may make the overall optimization costs inacceptable. In this paper, a computationally efficient simulation-driven DRA design is demonstrated. We present two DRA design cases with antenna geometry adjusted by means of surrogate-based optimization with the underlying low-fidelity model constructed from coarse-discretization EM simulations of the antenna of interest. The specific design approaches utilized here are space mapping and adaptively adjusted design specifications. For both examples, the optimized design is found at the computational cost corresponding only to a few high-fidelity simulations of the DRA.