PHI Lab at Bar-Ilan University studies what human movement and physiological signals reveal about health, motor function, and adaptation. Our research combines machine learning, computer vision, and multimodal sensing to support clinical assessment and personalized rehabilitation. We also investigate how robots learn from human experience and how people and robots adapt and solve physical tasks together.
Physical Human Intelligence
We develop multimodal AI methods to infer underlying motor and functional states from movement and physiology, aiming to detect impairment, characterize disease severity, and track functional change. Key questions include how to distinguish persistent motor impairment from temporary fatigue and how to identify compensatory movement strategies. Walking, balance, and reaching tasks provide settings for investigating functional differences that task completion alone may not reveal.
Personalized Rehabilitation
We investigate how rehabilitation robots and interactive games can adapt assistance, feedback, and challenge to each person, supporting motor learning, lasting recovery, and transfer to everyday activities. We ask when assistance improves immediate performance and when it supports learning that persists after support is reduced. Reaching tasks, balance exercises, and rehabilitation games provide settings for studying how gains transfer to independent everyday activity.
Learning from Human Experience
We study how robots can learn from the way people acquire skills—including exploration, errors, and strategy changes—combining demonstrations, language, and movement to transfer knowledge across bodies and tasks. By examining how human strategies evolve during practice, we aim to identify what a robot can learn from successful and unsuccessful attempts. A central question is how this experience can guide a robot facing an unfamiliar task or different physical constraints.
Adaptive Human–Robot Collaboration
We investigate how people and robots adapt together when tasks or conditions change, coordinating roles, sharing information, and recovering from unexpected events to solve physical problems jointly. We ask how assistance and role allocation affect both immediate coordination and each partner’s ability to adapt. Shared object-handling tasks provide a setting for examining how human and robot strategies change when familiar solutions become less effective.
Related publications
Selected publications by the principal investigator and collaborators.
- Bridge-Disease Curriculum Transfer Enables Smartphone-Based Balance Assessment in Rare Neurologic Disorders — IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2026.
- Diffusion-Augmented Spatiotemporal Graph Convolution for Clinical Gait and Motor Function Assessment — IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2026.
- Speech-to-trajectory: learning intuitive verbal guidance for robot motion — Intelligent Service Robotics, 2026.
- DiG-Net: Enhancing Human–Robot Interaction through Hyper-Range Dynamic Gesture Recognition in Assistive Robotics — ACM Transactions on Human-Robot Interaction, 2026.