A MOEA/D-IHPSO-based multi-objective optimization for dual-polarized omnidirectional antenna design

Abstract

A multi-objective evolutionary algorithm based on decomposition with an improved hybrid particle swarm optimization (MOEA/D-IHPSO) method is proposed, which aims to enhance the efficiency of automated optimization for antenna geometry. The optimization process includes two stages: optimization problem modeling and algorithm application. In the first stage, the topology optimization regions are selected according to the surface current distribution of the antenna, and a differential grid division strategy is employed to characterize the selected regions, ensuring the modeling accuracy of the key topology areas. By combining the encoded topological parameters of the selected regions with the key size parameters, a hybrid solution vector is formed and subsequently used to generate the initial population. In the second stage, the proposed framework uses new crossover and adaptive mutation operators to accelerate the process of reaching the target solution. Furthermore, MOEA/D-IHPSO incorporates a novel transfer learning-based recombination strategy that directly leverages historical optimization data without requiring surrogate model training, thereby intelligently guiding the search direction. To validate the effectiveness of the method, a dual-polarized omnidirectional antenna is optimized. Experimental results show that the optimization efficiency is improved by 31.8% and 26% compared to IBPSO and MOEA/D-GO algorithms, respectively.