Title: Towards safe, robust, and trustworthy robot autonomy: Infusing formal methods into learning-enabled autonomy Abstract: Advances in the field of artificial intelligence and machine learning have enabled robotic systems to operate in unstructured, uncertain, and novel environments, especially in safety-critical settings that involve interactions with humans. Yet as learning-enabled techniques such as deep learning continue to become more pervasive, it also becomes increasingly difficult to ascertain the performance and safety of such learning-enabled robotic systems, and explain their behavior. In this talk, I will discuss how we can leverage techniques from formal methods, namely Hamilton-Jacobi (HJ) reachability and Signal Temporal Logic (STL), to complement a learning-enabled robot autonomy stack and hence design safer and more robust robot behaviors. First, I will introduce HJ reachability theory and show how to incorporate HJ reachability analysis into a robot autonomy stack to produce minimally interventional safe planning and control strategies. Second, I will introduce STL and demonstrate how to leverage STL to incorporate human-domain knowledge into learning-based components of an autonomy stack and thus produce outputs that are more robust and interpretable. I will end the talk by showing deep connections between HJ reachability, STL, and machine learning and thus lay the groundwork for exciting future research directions. Bio: Karen Leung is a PhD candidate in Aeronautics and Astronautics at Stanford University. She works with Professor Marco Pavone in the Stanford Autonomous Systems Lab. Her research harnesses advances in learning-empowered robot autonomy and unites them with the rigor and assurances provided by formal methods to develop powerful yet safe and trustworthy autonomous systems. Her research is geared towards achieving human trust in learning-enabled robot autonomy, starting from the algorithmic foundations of safe robot decision making and control, and incorporating further refinement through learnings from practical deployment. Karen received a combined degree in Aerospace Engineering and Mathematics from the University of Sydney in 2014. She was a recipient of the Qualcomm Innovation Fellowship and the Robert H. Cannon Jr. department fellowship for outstanding contributions and achievements in 2018, and was co-chair in the 2018 nationwide Women in Aerospace Symposium. |