Our Research Our work lies at the intersection of control theory, machine learning, and formal methods. Here are some of the areas we're currently working on. Home | REALM | Our Research Research Themes Certified learning for control Before deploying deep learning-based methods for control, we need to understand the safety of these controllers. We develop tools to make learned controllers safer and generate data-driven proofs (certificates) of correctness. Automated testing, verification & design optimization As autonomous systems becomes more complex, they can start to fail in unexpected ways. We develop tools that help engineers discover, understand, and mitigate these failure modes. Safe human-robot interaction and planning We develop techniques for ensuring both the physical safety and intelligent responsiveness of robot agents in responding to complex human instructions, enabling robots to successfully work together with humans. LLM-based human-machine interaction We merge Foundation Models with symbolic computing to enhance general planning and decision-making in robots, covering tasks like drone/manipulator motion planning and applications like travel planning and web agents.
