Welcome to the Intelligent Robotics and Interactive Systems (IRIS) Lab!

Robots today can see, move, and learn—but they still struggle with what humans do effortlessly: touch. Picking up a fragile egg, rotating a key, making you a coffee—these demand a fluid dance of fingers and forces that robots find surprisingly hard. At the IRIS lab, we are developing

Dexterous manipulation with physics-grounded intelligence. We enable multi-fingered robotic hands to manipulate objects with human-like skill. Our approach fuses top-down learning with bottom-up contact physics—differentiable simulation, optimization, and planning that respect the realities of touch. The result: robots that learn from minimal data and generalize broadly.

Robots that learn from you, safely. Teaching a robot shouldn’t require programming or thousands of demos. We design transparent and structured learning methods that let robots acquire skills from natural interactions—sparse demonstration, physical nudge, a simple preference. Our methods come with certifiable guarantees: safety and alignment with your intent.

Together, we’re building toward capable, trustworthy robots that act skillfully and collaborate seamlessly with people.

               

Recent Updates