Inverse Design Method of Antenna Based on Generative Artificial Intelligence

An antenna inverse design scheme based on generative artificial intelligence (AI) is proposed. The generative AI uses a diffusion model and incorporates a transformer model to output a series of innovative and practical antenna structures according to the input antenna requirements. To select the optimal structure from a large set of candidates, we further incorporate the Characteristic Mode Analysis (CMA) method into the Bilinear Convolutional Neural Networks (BCNN) model for evaluation and selection. The antenna inverse design scheme cleverly combines AI with the CMA method, which greatly simplifies the tedious process of traditional antenna design. By simply inputting the antenna requirements, the optimal antenna structure, distinct from existing datasets, can be rapidly obtained. Since the output is an antenna structure, it is not limited to a specific frequency band, thus offering design flexibility for cross-band adjustments. This letter presents case studies on single-band, dual-band, and wideband circularly polarized microstrip antennas to validate the practicality and effectiveness of the proposed scheme.