Robot-assisted rehabilitation: Collaborating a smart avatar with an AAN controller.

Area of Science:

• Robotics
• Rehabilitation Engineering
• Biomedical Engineering

Background:

• Current robot-assisted rehabilitation lacks real-time personalization, limiting intervention effectiveness.
• Physiological Functional Capacity (PFC) is often not integrated into assistive technology design.
• Existing Assist-as-Needed (AAN) strategies require enhanced subject-specific adaptation.

Purpose of the Study:

• To develop an adaptive Assist-as-Needed (AAN) controller integrated with a Smart Avatar for personalized rehabilitation.
• To enhance real-time personalization by incorporating a Smart Avatar that learns and mimics patient capabilities.
• To improve energy efficiency in robot-assisted therapy through modified reference trajectories.

Main Methods:

• Designed a Reinforcement Learning (RL) based Inverse Dynamics model to predict subject involvement.
• Developed a Smart Avatar to learn and mimic real-time patient capabilities.
• Integrated an Energy Map into the controller to optimize reference trajectories for energy efficiency.

Main Results:

• The AAN controller, guided by the Smart Avatar, achieved low trajectory tracking errors (0.01-0.04 rad) in wrist motions.
• Human torque estimation by the Smart Avatar enabled optimized robot torque delivery.
• Analysis revealed subject- and axis-dependent interaction torque differences, confirming adaptive assistance feasibility.

Conclusions:

• The developed Smart Avatar-integrated AAN controller demonstrates feasibility for personalized robot-assisted rehabilitation.
• Real-time adaptation based on learned patient capabilities enhances control accuracy and energy efficiency.
• This approach offers a promising direction for optimizing assistive technology in rehabilitation settings.