Physics-Informed AI for Wireless Communication and Sensing

Deep learning models encounter fundamental challenges when directly applied to wireless systems, including limited interpretability, low data efficiency, and poor adaptability in dynamic environments. To overcome these challenges, this extended abstract presents my research on embedding electromagnetic physics into AI frameworks to create physics-informed models for wireless communication and sensing. These hybrid models leverage the structure of physical laws alongside data-driven learning, leading to substantial improvements in accuracy, generalization, interpretability, and robustness across complex scenarios. My work demonstrates how this methodology reforms wireless systems across multiple key applications: channel prediction, indoor localization, antenna design, and hardware fingerprinting. Together, these efforts establish a new paradigm for next-generation wireless system design.