Generative Adversarial Network-Based Design of Dielectric Resonator Antenna for mmWave 5G Applications

Characterized by low conduction loss and high radiation efficiency, dielectric resonator antennas (DRA) have been considered as a promising antenna type for the applications of 5G technologies at 60 GHz band. However, the design of the DRA mainly depends on the intuitive reasoning and trial-and-error process, which is time-consuming and resource-demanding. To address this challenge, we developed a generative adversarial network (GAN)-based approach to assist the design of DRA structures. This GAN model incorporates a simulator and generator for the forward prediction and inverse design, respectively. A paired dataset of geometric shapes and S11 parameter was fed into the GAN model to train the neural network. During the training process, the distribution of this dataset was captured by the end-to-end GNA model. The final model is capable of generating new DRA structures according to the desired S11 parameter.