Safe and Intelligent Autonomy Lab

Welcome to the Safe and Intelligent Autonomy (SIA) Lab! SIA Lab is part of the Department of Aeronautics and Astronautics at Stanford. Our lab develops principled methods for ensuring the safety of learning-enabled autonomous systems.

Modern robots increasingly rely on machine learning, foundation models, and data-driven decision-making to perceive, plan, and act in complex environments. These capabilities create tremendous opportunities, but also introduce new safety challenges: failures can arise from imperfect models, distribution shift, brittle policies, untested scenarios, and unexpected interactions with the real world.

Our central goal is to build the safety lifecycle for learning-enabled autonomy. We develop theory, algorithms, and computational tools that help autonomous systems discover failures before deployment, intervene safely during deployment, and improve from failures and near-misses after deployment.

The technical backbone of our work is principled safety: reachability analysis, control theory, planning, verification, and learning-based safety representations. We use these foundations to design safer policies, evaluate their limitations, deploy them with runtime safeguards, and continually adapt them to new safety risks.