Robot learns from therapists to support stroke rehabilitation

Strokes leave 5 million people with long-term impairments each year, while a growing shortage of physical therapists limits access to care. A robotic rehabilitation system aims to address that gap by learning from human therapists and adjusting the assistance it provides during exercises.

Johannes Lachner, a former MIT postdoc now serving as an assistant professor at Purdue University, developed the system with Noah Geiger, a former visiting student at MIT. It combines generative AI and real-time force feedback to support movements such as reaching, lifting the arms, and performing tasks of daily living. The goal is to extend therapists’ reach rather than replace them.

Initial robotic experiments have demonstrated the approach’s feasibility, according to Lachner. The next phase aims to develop therapist-specific models and evaluate them in a long-term clinical study involving the same patients who previously received manual therapy. Lachner also envisions possible applications in contact-rich industrial settings and collaborative workspaces.