Element-concurrent multi-feature surrogate differential evolution algorithm for efficient design of low scattering metasurfaces

Conventional approaches for the design of low scattering metasurfaces are mainly intuition-inspired. Therefore, they are unable to identify truly optimal solutions. In this article, an element-concurrent multi-feature surrogate differential evolution (EC-MFSDE) algorithm is proposed for efficient design of low scattering metasurfaces. Firstly, a good point set sampling (GPSS) method is employed to generate the initial sampling sets that can be more uniformly scattered in the search space. Secondly, an element concurrent optimization strategy is adopted to simultaneously optimize the structure parameters of each element, which can improve the design freedom. Thirdly, the element and fitness databases are built for training surrogate models and generating parent populations, respectively. Finally, by using these two databases, a multi-feature surrogate technique is designed to extract and predict the EM response accurately. Three metasurfaces are redesigned to validate effectiveness of the proposed algorithm. Simulation results confirm that the EC-MFSDE algorithm is an excellent candidate for the design of low scattering metasurfaces.