Research Thrusts
Robot Foundation Models
Generalist robot policies
We develop the learning and reasoning foundations behind generalist robot policies — making vision-language-action and world models dependable through test-time scaling, reasoning, and physically grounded simulation and modeling.
Embodied AI Agents
From digital to physical
We extend the orchestration, planning, and world knowledge of modern agentic systems to physical robots, grounding them in multi-modal perception and online interaction.
Physical AI Safety
Safe and reliable autonomy
Our research integrates monitoring, guardrailing, and alignment techniques that keep embodied systems safe, predictable, and reliable in the real world.
Data Flywheels for Robotics
Self-improving autonomy
We design automated evaluation and data pipelines that close feedback loops and enable continuously advancing robotic systems.
Optimal & Learning-based Control
Operate at the limits
We look at the complementary nature of optimization-based and learning-based control and planning methods. At their interface, we can bring the speed, interpretability, and constraint satisfaction of optimization to the broad capabilities and generalization of machine learning.
Applied Autonomy
Deploy in high-impact domains
We deploy autonomy in the field — spanning space robotics, legged and humanoid systems, autonomous vehicles, and manipulation — where operating in the world's most challenging environments grounds and drives our core research.
