Robust surrogate-based optimisation of planar metamaterial structures

In this paper, we presented a novel and efficient automated implementation of an ASM synthesis approach for planar metamaterial structures. Satisfactory results are achieved using our implementation after only a few fine model evaluations, eliminating the computational expense of direct optimisation of the high-fidelity full-wave EM model. At the core of this approach is a quasi-Newton model-trust region optimisation algorithm that provide the robustness required to implement an ASM process by combining the global convergence properties of steepestdescent and the fast local convergence of quasi-Newton´s methods.