Published: Apr 15, 2026 by Daning Huang
We hosted Dr. Fotis Kopsaftopoulos from Rensselaer Polytechnic Institute for a CAV talk via the AI/ML technical group.
Abstract: Future intelligent systems will be able to “feel,” “think,” and “react” in real time by leveraging multi-modal sensing and advanced data-driven frameworks, achieving accurate and robust state awareness and self-diagnostic capabilities. This talk introduces a unified stochastic framework for state awareness across diverse engineering domains, from “fly-by-feel” aerial vehicles inspired by avian flight to structural health monitoring (SHM) and additive manufacturing (AM) process monitoring. We discuss how stochastic modeling, advanced functional time series analysis, and machine learning can provide insight into evolving system states, tackling both forward and inverse problems, under uncertain conditions. Recent advances in our state awareness framework are presented, including stochastic system identification methods, Physics-Informed Neural Networks (PINNs), and Time Series Foundation Models (TSFMs), highlighting how time series representations serve as the connective thread across these methodologies for self-aware, self-diagnostic systems. In the aerospace domain, we show how embedded sensing and identification frameworks move beyond structural integrity assessment, based on vibration and acousto-elastic guided wave data, to inform flight state estimation through fusion of structural and aerodynamic feedback, enabling detection of critical phenomena such as stall, flutter, or sudden flow changes. In AM, we demonstrate how the same principled framework enhanced by PINNs extends to online process monitoring for metal AM methods. Experimental results are demonstrated from prototype fixed-wing aircraft, full-scale UAVs, and multicopters spanning flight tests and wind tunnel/laboratory experiments along with AM case studies. Ultimately, these advancements represent a transformative step towards self-aware intelligent systems.