Soft Prescribed Performance-Based Reinforcement Learning Control for a PAM-Actuated Rehabilitation Exoskeleton

Area of Science:

• Robotics
• Control Systems
• Biomedical Engineering

Background:

• Rehabilitation exoskeletons require stable and safe operation for effective training.
• System degradation can conflict with performance constraints, posing safety risks.

Purpose of the Study:

• To develop a novel control method for rehabilitation exoskeletons that ensures both high accuracy and safe operation.
• To address the challenge of performance degradation in rehabilitation robots.

Main Methods:

• A soft prescribed performance (SPP) reinforcement learning (RL) control strategy was developed.
• A tunnel-type prescribed performance function and dynamic adjustment of soft boundaries were employed.
• An actor-critic (AC) structure was used to manage unknown disturbances, with theoretical stability analysis.

Main Results:

• The proposed SPP-RL method ensures high accuracy and safe operation by dynamically adjusting performance constraints.
• Theoretical analysis confirmed the closed-loop system's stability.
• Experiments on an upper-limb exoskeleton robot demonstrated the method's effectiveness and robustness.

Conclusions:

• The developed SPP-RL control method effectively enhances the safety and accuracy of rehabilitation exoskeletons.
• The approach is robust and adaptable to system degradation, crucial for rehabilitation training.