“Assured Autonomy in an Open Dynamic World with Integrated Formal Methods”
Tuesday, May 11 at 1:00pm
Email email@example.com for Zoom Info
To enable large-scale adoption of autonomy, establishing trust and assurance is crucial. This task is nontrivial because of the increasing need for modern systems to interact with an open and dynamic world.
The challenges in dealing an open world involve two main cases: An adversarial environment with asymmetric information; and a stochastic environment with unknown dynamics. The first case is commonly encountered in defense and security applications. One such problem is on synthesizing automated cyber defense, where asymmetric information is introduced by the deployment of honeypots and other deception mechanisms. Integrating formal methods with game theory, I will show how to leverage high-level information for strategic advantages and correctness guarantees respecting rich temporal logic specifications. The second case is encountered in exploration missions or human-robot interaction, for which reinforcement learning becomes indispensable for adaptive systems. I will discuss how logical composition and task structure can be used for improving sample efficiency and learning transferable skills. Lastly, I will discuss the coupling between cyber defense and physical interaction and present our ongoing work on cyber-physical security and preference-aware decision-making.
Dr. Jie Fu is an assistant professor with the Department of Robotics Engineering from Worcester Polytechnic Institute, with affiliation to the Department of Electrical and Computer Engineering. She earned her Ph.D. degree in Mechanical Engineering from the University of Delaware. She held post-doc research positions at the University of Pennsylvania. Her research focuses on theory and algorithms for synthesizing correct-by-construction autonomous systems through integrated formal methods, learning, control, and game theory. Her research is sponsored by NSF, AFOSR, ARO, and DARPA. She received the AFOSR YIP award in 2020.