AE 590 Seminar Speaker: Inseok Hwang - Toward Trustworthy Human-aware Cyber-Physical-Human Systems (CPHS)

Abstract
In this talk, I present our lab’s recent research on high-assurance autonomy for Cyber-Physical-Human systems (CPHS), with a particular focus on human-aware autonomy. Despite recent advances in autonomous systems and AI, many real-world applications still require close interaction with humans. Accordingly, one needs to fully accommodate human behavior that is inherently uncertain, variable, and often difficult to predict, depending not only on the environment but also on individual characteristics and cognitive states. This yields a need for a method to formally guarantee the safety of such human-in-the-loop systems. Our team has been addressing these challenges in CPHS from multiple perspectives, including control-theoretic and data-driven approaches. I will discuss our efforts, including cognitive-aware autonomous teammates that adapt their assistance to human users, skill-level inference for selecting appropriate training support, and data-driven safety analysis based on reachability. These studies aim to develop autonomous systems that can effectively account for human uncertainty while maintaining reliability. In addition, our recent work on the formal safety analysis using signal temporal logic (STL) will be presented. In this work, we explored how formal methods can be used to verify the teamwork between humans and Large Language Model (LLM)-based agents. Together, these efforts highlight a path toward CPHS that are not only capable of safely interacting with humans, but also adaptive, reliable, and formally verifiable.

Bio
Dr. Hwang is the Paul Stanley Professor in the School of Aeronautics and Astronautics at Purdue University. He is a member  of Purdue's Signature Area for System-of-Systems research. He earned his Ph.D. degree, specialized in the area of multiple-vehicle control and its application to air traffic control using hybrid systems approach, in the Department of Aeronautics and Astronautics at Stanford University. He was a member of the Hybrid Systems Laboratory at Stanford University. He was a full time instructor in the Department of Aerospace Engineering at Korea Air Force Academy from 1994 to 1997. He is an Associate Fellow of the American Institute of Aeronautics and Astronautics (AIAA) and is a member of the Institute of Electrical and Electronics (IEEE) Control Systems Society and Aerospace & Electronic Systems Society. He is currently an Senior Editor of the IEEE Transactions on Aerospace and Electronic Systems.

Dr. Hwang's research has been strongly motivated by difficult and interesting practical problems such as controlling multiple-vehicle systems. Controlling multiple-vehicle systems is one of the most important and challenging aspects of modern system theory and practice. Control of such systems involves the analysis of multiple dynamical systems which have inherently decentralized structures. The motions of vehicles have to be coordinated in such a way that the vehicles achieve their goals without conflicts between them. This requires path planning (computing optimal trajectories of vehicles from starting positions to destinations) and conflict detection and resolution. Path planning and conflict detection and resolution require information about individual vehicles, and therefore communication between vehicles for sharing this information is important. Multiple-vehicle systems encompass a variety of applications, including groups of Unmanned Aerial Vehicles (UAVs), Satellites, and mobile robots; ad-hoc sensor networks; air traffic control.