Evaluation of a machine-learning-driven active-passive upper-limb exoskeleton robot: Experimental human-in-the-loop...

Area of Science:

• Robotics
• Biomechanics
• Human-Robot Interaction

Background:

• Designing effective exoskeleton controllers is challenging due to complex human-robot interactions, sensor/actuator selection, and communication issues.
• Existing methods for evaluating exoskeleton performance can be time-consuming and may not fully capture user experience.

Purpose of the Study:

• To develop a test framework for evaluating a novel active-passive shoulder exoskeleton.
• To create and validate a surface electromyography (sEMG)-based human-robot cooperative control method for intuitive exoskeleton operation.

Main Methods:

• A hierarchical control strategy integrating sEMG-based intention estimation, strength regulation, and actuator control was implemented.
• Shoulder joint elevation experiments were conducted comparing active-passive, fully passive, and fully active exoskeleton control.
• Performance was assessed using post-test surveys, load tolerance duration, and inverse dynamic simulation of human torque, power, and metabolic energy expenditure.

Main Results:

• The active-passive exoskeleton demonstrated a 50% reduction in required user muscular activation torque compared to fully passive control.
• Fatigue indicators were reduced by a factor of 3 in the active-passive condition relative to the fully passive condition.
• The sEMG-based control method effectively translated wearer movement intentions into exoskeleton assistance.

Conclusions:

• The developed active-passive shoulder exoskeleton and its sEMG-based control system offer significant advantages in reducing user effort and fatigue.
• The novel test framework provides a robust method for evaluating exoskeleton performance and control strategies.
• This research contributes to the advancement of more intuitive and efficient human-robot cooperative systems for wearable assistive devices.