Human-in-the-Loop Modeling and Bilateral Skill Transfer Control of Soft Exoskeleton

Area of Science:

• Robotics
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Soft exoskeletons offer compliant assistance compared to rigid robots.
• Hemiplegic patients require effective rehabilitation strategies for limb recovery.
• Existing solutions may lack high-strength, variable-stiffness actuation.

Purpose of the Study:

• To develop a soft exosuit with twisted string actuators (TSAs) for hemiplegic patients.
• To create an adaptive impedance controller for mirror training between healthy and impaired limbs.
• To optimize robotic assistance using reinforcement learning (RL).

Main Methods:

• Formulated an analytic model of the TSA.
• Developed a human-exosuit coupled dynamic model.
• Designed an adaptive impedance controller for skill transfer (movement trajectory and stiffness).
• Implemented a reinforcement learning algorithm to optimize impedance parameters.

Main Results:

• Demonstrated the effectiveness of the developed exosuit and control strategy.
• Showcased the superiority of the proposed method in experimental trials.
• Validated the adaptive impedance controller's ability to facilitate mirror training.

Conclusions:

• The soft exosuit with TSAs provides high-strength, variable-stiffness actuation.
• The adaptive impedance controller and RL-based optimization enhance rehabilitation for hemiplegic patients.
• This approach offers a promising solution for assistive and rehabilitative robotics.