AI-Driven Electromagnetic Co-Design: From Combinatorial RF Topology Optimization to Physics-Aware Surrogate Modeling
October 1, 2026 · 5:00 AM - 6:00 AM @ Moratuwa, Department of Electronic and Telecommunication Engineering
Description
This seminar presents an AI-driven framework for automated electromagnetic co-design, spanning RF topology generation, multifunctional hardware structures, and fast surrogate modeling. First, binary-matrix optimization is used to explore non-intuitive microstrip resonator and inductor geometries through openEMS simulation and evolutionary optimization. Second, this concept is extended to QR-coded microstrip antennas, where digital identity and RF behavior are jointly optimized for IoT and hardware verification. Finally, 50 neural-network architectures are benchmarked for EM surrogate modeling, revealing the importance of spatial inductive bias and physics-informed learning. Together, these works demonstrate a pathway toward scalable, intelligent, and automated RF design.