Stroke is a leading cause of long-term disability, leaving about 5 million people worldwide with lasting impairments each year. At the same time, the supply of physical therapists is not keeping up with demand, making it difficult for many patients to get the rehabilitation they need. Researchers are now testing a robotic system designed to ease that pressure by learning from human therapists.
The system, developed by Johannes Lachner, now at Purdue University, and Noah Geiger, a former visiting student at MIT, combines generative AI with real-time force feedback. It observes how therapists guide patients through exercises such as reaching and lifting, then provides assistance calibrated to each patient's needs. The developers say the goal is not to replace therapists but to extend their reach, allowing more patients to receive consistent support.
Early experiments have shown the approach is feasible, according to Lachner. The next phase will focus on building therapist-specific models and running a long-term clinical study with patients who previously received manual therapy. The researchers also see potential uses beyond rehabilitation, including contact-rich industrial tasks and collaborative workspaces.