Prior Knowledge-Guided Deep Learning-Enabled Metantenna Optimization and Synthesis (Invited)

This paper reviews recent advances in deep learning (DL)-enabled techniques for optimizing and synthesizing metamaterial-based antennas (metantennas). Expanding the degrees of freedom (DoF) of design of metamaterials and metasurfaces creates new opportunities for innovative metantenna designs by the introduction of novel physical properties. However, the increased DoF also increases significant complexity of the design process. To tackle this challenge, prior-knowledge-guided deep learning-enabled (PK-DL) optimization and synthesis methods have been proposed and developed. These methods integrate prior knowledge into deep learning algorithms, such as artificial neural networks (ANNs) and conditional deep convolutional generative adversarial networks (cDCGANs), to guide and accelerate metantenna design processes. Recent progress demonstrates that AI algorithms applied to antenna design open new opportunities for optimizing and synthesizing metantennas with enhanced performance and multifunctional capabilities.