Our research spans physical AI, robot learning, computer vision, and multimodal architectures. We focus on frameworks for sim-to-real transfer and zero-shot motor control, enabling machines to adapt and perform complex tasks in unpredictable environments.
Key initiatives include spatial reasoning, manipulation, and autonomous navigation. We also develop human-robot interaction models that combine safety bounds with adaptive force regulation, ensuring robust and intuitive physical collaboration.
Grounding abstract neural models in physical conservation laws.
True embodied reasoning requires spatial perception and mechanical action to converge in real time, bridging digital intelligence with the tangible constraints of the physical world.
Macro optical array data visualizing real-time spatial point clouds during closed-loop motion control testing.
We invite academic partners and researchers to collaborate on advancing the frontiers of embodied perception and autonomous control systems.