Adaptive Learning based Upper-Limb Rehabilitation Training System with Collaborative Robot.

Area of Science:

• Robotics
• Rehabilitation Engineering
• Human-Computer Interaction

Background:

• Traditional rehabilitation requires specialized centers and devices, limiting accessibility and increasing costs.
• Home-based training offers a more accessible and cost-effective alternative for patients with motor disabilities.

Purpose of the Study:

• To investigate the feasibility of using a general-purpose collaborative robot for multi-purpose upper-limb rehabilitation training at home.
• To develop and evaluate a robot system that integrates human motor feedback and preferences for personalized training.

Main Methods:

• Developed a system using a generic robot arm integrated with surface electromyography, force/torque sensors, and RGB-D cameras via the Robot Operating System.
• Employed imitation learning to adapt expert training trajectories to individual subject capabilities.
• Simulated system control and training effects, followed by real-world evaluation with human subjects.

Main Results:

• The developed system successfully performed gross and fine motor skill training using a gripper-based end-effector.
• Simulations in Gazebo and Open-Sim validated system control and predicted training effects.
• Human subjects demonstrated superior training outcomes with the collaborative robot system compared to specialist-assisted rehabilitation.

Conclusions:

• Collaborative robots can be effectively utilized for safe and convenient in-home motor rehabilitation training.
• The system offers a viable alternative to specialized rehabilitation robots, improving treatment accessibility and potentially reducing healthcare costs.