Surrogate Assisted Multi Objective Optimization of a 2.4 GHz Yagi Antenna Using Gaussian Process Regression and MOPSO

Abstract

This paper presents a surrogate‑assisted multi‑objective optimization approach for a 2.4 GHz Yagi–Uda antenna using Gaussian Process Regression (GPR) and Multi‑Objective Particle Swarm Optimization (MOPSO). The antenna performance is evaluated in terms of maximum gain and minimum Voltage Standing Wave Ratio (VSWR). Three key physical parameters—inter‑element spacing (gap), director length, and element diameter—are parametrically swept using CST Microwave Studio to generate training data. GPR models are employed as surrogate fitness estimators to approximate the nonlinear relationship between antenna geometry and performance metrics. These surrogate models are integrated into the MOPSO framework to efficiently explore the design space and obtain Pareto‑optimal solutions. The results demonstrate that the proposed method significantly reduces the computational cost while effectively identifying optimal trade‑offs between gain and VSWR. A maximum gain of 9.4 dBi with a VSWR of 1.18 is achieved at 2.4 GHz, validating the effectiveness of the proposed approach for practical WiFi antenna design.