Optimization of an Irregular Slot Antenna Based on Conditional Generative Adversarial Networks

In this paper, to address the problem of multilayer perceptron (MLP) with weak generalization ability, conditional generative adversarial networks (CGAN) is proposed to learn the complex nonlinear relationship between antenna structure parameters and performance parameters. CGAN in combination with Genetic Algorithm (GA) can be used to achieve antenna optimization quickly. The simulated impedance bandwidth obtained by the manual design of an irregular slot antenna is 45.9% (2.28-3.61 GHz) and the optimized impedance bandwidth is 49.66% (2.24-3.72 GHz). The experimental results show that the predicted data basically match the HFSS simulation data and the optimization objective is well accomplished. The method proposed in this paper speeds up the antenna optimization process and CGAN performs better in the optimization compared to MLP. The effectiveness and reliability of the method is demonstrated.