Research
Autonomous systems are increasingly powered by machine learning, foundation models, and data-driven decision-making. These methods enable robots to operate in complex environments, but they also introduce new safety challenges: brittle policies, imperfect models, distribution shift, untested scenarios, and unexpected interactions during deployment.
At SIA Lab, we develop lifecycle safety methods for learning-enabled autonomy. Our goal is to help autonomous systems be designed with safety in mind, monitored and corrected during deployment, and improved from failures and near-misses over time. Across these stages, our work is grounded in principled safety foundations, including reachability analysis, control theory, planning, verification, and learning.
Our research is organized around four themes.
