My primary research interests lie somewhere in the intersection of machine learning, classical control theory, and formal methods. We have developed extensive methodologies in each of these fields for approaching humanity's desire for autonomous systems, but we currently lack robust ways to provide performance guarantees for autonomous, safety-critical systems. To bring us closer to provably safe autonomous systems, I focus on ways to use machine learning to learn high-level human goals from demonstrations in a format (automata) that is amenable to formal control synthesis. I am also investigating ways to bound uncertainty on the learned human objectives, with the goal being to create principled, auditable, and safe end-to-end learning based decision-making systems. Ultimately, we need more principled and guaranteed design methodologies than many "black-box" methods in use today if we want to realize trusted autonomy for safety-critical systems in our society.
If you want to check out what I'm up to, check out my github page for all the latest and greatest codings and algorithms!
Research Keywords: Robotics, Formal Control Synthesis, Artificial Intelligence