Research
We work on physical AI to understand and generate human physical interaction abilities.
As a guiding principle, we use advanced differential-geometric methods, resting on coordinate invariance, stability, safety, and computational efficiency.
Research thrusts
Learning from physical therapists
We monitor physical and occupational therapists during physical rehabilitation, then train robotic systems that can emulate them. Our goal is that more patients can be treated, and underserved communities can benefit.
Neural-driven prostheticsTraining prosthetic devices
Prosthetics and brain-machine interfaces focus almost entirely on motion generation. We want to let patients train their own prosthetic device with their intact hand, giving them the ability to physically interact with the world the way a healthy human hand does.
Human Motor ControlThe structure of physical interaction
What we learn from building these systems, and from watching patients recover, feeds back into the open questions of motor neuroscience. Our guiding research question: how can people excel at physical interaction despite slow signaling and neuromuscular noise?