AI-Enhanced Motor Power Analysis for Joint Kinematics Prediction and Fault Detection in Rehabilitation Systems

Area of Science:

• Biomedical Engineering
• Robotics
• Rehabilitation Technology

Background:

• Post-stroke recovery necessitates effective motor function rehabilitation.
• Assistive robotic systems are crucial for physiotherapy but require high safety standards.
• Accurate joint angle measurement is vital for precise feedback in rehabilitation devices.

Purpose of the Study:

• To introduce a novel fault-alert exoskeleton wheelchair for upper and lower limb rehabilitation.
• To enhance the safety and accuracy of robotic-assisted physiotherapy for post-stroke patients.
• To develop a robust joint angle prediction system to overcome sensor limitations.

Main Methods:

• A redundant mechanism fault-alert exoskeleton wheelchair design was developed.
• A machine learning (ML) model was employed for joint angle prediction using power analysis from current sensors.
• A Collective Generator Network (CGN) was utilized to synthesize power profiles based on user physicality.

Main Results:

• The proposed ML and CGN models effectively detect electrical, mechanical, and encoder issues in the rehabilitation system.
• The system enhances the accuracy of joint angle measurement, crucial for precise rehabilitation feedback.
• Freely available ML, CGN models, and datasets encourage wider adoption in research.

Conclusions:

• The integration of joint kinematics prediction via motor power analysis and CGN enhances rehabilitation system safety.
• Subtle fault variations within the system can be detected, improving patient care.
• This approach offers a significant advancement in ensuring reliable and effective robotic-assisted physiotherapy.