Design Space Reduction for Expedited Multi-Objective Design Optimization of Antennas in Highly Dimensional Spaces

Abstract

A surrogate-based technique for efficient multi-objective antenna optimization is discussed. Our approach exploits response surface approximation (RSA) model constructed from low-fidelity antenna model data (here, obtained through coarse-discretization electromagnetic simulations). The RSA model enables fast determination of the best available trade-offs between conflicting design goals. The cost of RSA model construction for multi-parameter antennas is significantly lowered through initial design space reduction. Optimization of the response surface approximation model is carried out by a multi-objective evolutionary algorithm (MOEA). Additional response correction techniques are subsequently applied to improve selected designs at the level of high-fidelity electromagnetic antenna model. The refined designs constitute the final Pareto set representation. The presented multi-objective design approach is validated using three examples: a six-variable ultra-wideband dipole antenna, an eight-variable planar Yagi-Uda antenna, and an ultra-wideband monocone with 13 design variables.