Hosted by Lisa Kusch

Abstract

Recent advances in machine learning and artificial intelligence are opening new significant opportunities for the design, modeling, monitoring, and operation of physical systems as next-generation electromagnetic and electronic systems in applications as communications, computing, sensing, and imaging. This seminar will present research results for electromagnetic design, with applications as metasurfaces, defective ground plane filters, and THz sensors. These techniques allow building models that can be used for efficient and accurate design tasks (e.g., optimization and uncertainty quantification tasks). Some results where physical insight and equivalent circuits representations are used instead of more general black-box descriptions will be shown. Connections with topics as system identification, surrogate modeling, and model order reduction will be illustrated.

The seminar will also present multiscale simulation and design methodologies, focusing on hybrid approaches that combine full-wave electromagnetic solvers with ray-tracing techniques. Such methods are particularly relevant for the design and analysis of metalenses and metasurface-enabled propagation environments, where accurate local electromagnetic modeling must be integrated with large-scale wave propagation analysis. Applications of metasurfaces in fluorescence endoscopy imaging and lab brain tissue validation will be shown.

Finally, the seminar will also discuss automated characterization and monitoring techniques including compressed-sensing-based fault identification in antenna arrays, embedded element pattern characterization methods, and adaptive behavioral modeling of power amplifiers from selectively acquired measurement data. The areas of advanced modeling approaches for electromagnetic spectrum anomaly detection and spectrum intelligence, and hardware-software codesign will also be briefly addressed.