Advancing Physical AI.
Research themes across world models, robot learning, embodied intelligence, spatial reasoning, control and manipulation.
World Models
Persistent representations of geometry, semantics, dynamics and uncertainty for physical reasoning.
Robot Learning
Learning behaviors and representations from interaction with physical environments.
Embodied Foundation Models
General-purpose models that connect perception, language, reasoning and robot action.
Spatial Intelligence
Understanding geometry, relationships, memory and change across real environments.
Generalization
Transferring useful behavior across objects, tasks, environments and robot embodiments.
Adaptive Control
Real-time execution that responds to physical uncertainty and change.
Manipulation
Dexterous and general interaction with objects, tools and environments.
Simulation-to-Real
Bridging learned behavior from simulated environments into real-world operation.
Continuous Learning
Improving models and behavior from deployment experience.