Identification of muscle-activation-dependent human-exoskeleton coupling parameters

Area of Science:

• Robotics
• Biomechanics
• Human-Machine Interaction

Background:

• Understanding human-exoskeleton interaction dynamics is crucial for effective exoskeleton design.
• Accurate prediction of coupling parameters is essential for improving control and performance.
• Existing models often lack the ability to adapt to varying muscle activation levels.

Purpose of the Study:

• To propose a muscle-activation-dependent human-exoskeleton model.
• To predict human-exoskeleton coupling parameters for enhanced dynamic studies.
• To establish a novel model for complex human-exoskeleton interaction scenarios.

Main Methods:

• Developed a new experimental platform for human-exoskeleton coupling analysis.
• Collected surface electromyographic (EMG) signals from 20 volunteers to represent muscle activation.
• Utilized a convolutional neural network (CNN) with six EMG features (MAV, MAVSLP, WL, WAMP, VAR, AR) to predict coupling parameters.

Main Results:

• The CNN model successfully predicted human-exoskeleton coupling parameters.
• Sensitivity analysis identified Auto Regressive (AR) coefficients, Mean Absolute Value (MAV), and Variance (VAR) as key determinants.
• A strong correlation was found between coupling stiffness and both MAV and VAR.

Conclusions:

• The developed muscle-activation-dependent model enables prediction of coupling parameters in complex scenarios.
• The novel experimental platform and modeling approach advance the study of human-exoskeleton dynamics.
• This work provides a foundation for more adaptive and intuitive exoskeleton control systems.