Research Statement

My research investigates how robots can use contact to understand the physical world. Many important properties of a scene—including hidden geometry, object relationships, and task-relevant constraints—cannot be reliably inferred from passive observation alone. Instead, robots must interact with their environment, using contact as a source of information to reduce uncertainty and reveal otherwise unobservable structure.

To study this problem, I develop simulation-based inference methods that treat physics simulation as a generative model of robot-environment interaction. By combining visual and tactile observations with probabilistic reasoning, robots can maintain beliefs over possible world states and update those beliefs through interaction. This approach enables scene understanding under uncertainty while naturally incorporating physical constraints and contact dynamics.

My research lies at the intersection of robot perception, probabilistic inference, active sensing, and contact-rich manipulation. The long-term goal is to develop interactive physical agents that understand their environment by reasoning about how actions, observations, and contact are connected through the underlying physics of the world.