sEMG-based joint force control for an upper-limb power-assist exoskeleton robot.

Area of Science:

• Robotics
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Existing exoskeleton control often relies on position control, which may not fully replicate natural human movement.
• Surface electromyogram (sEMG) signals offer a direct measure of user's muscle activation and motion intent.
• Power-assist exoskeletons require intuitive and responsive control for effective human augmentation.

Purpose of the Study:

• To develop and evaluate two novel sEMG-based force control strategies for power-assist exoskeleton arms.
• To enable the exoskeleton to provide more human-like assistance by controlling joint torques based on muscle signals.
• To improve the efficiency and simplicity of motion recognition and control in exoskeleton systems.

Main Methods:

• Investigated two distinct sEMG-based force control strategies for exoskeleton joint torque generation.
• Method 1: Estimated agonist and antagonist muscle forces to derive joint torque.
• Method 2: Employed linear discriminant analysis (LDA) classifiers for motion type recognition, combined with estimated muscle force for torque control, with a unique classifier per joint.

Main Results:

• The proposed force control methods enable the exoskeleton to mimic human-like behavior for enhanced assistance.
• The LDA-based method with individual joint classifiers simplified the recognition process and reduced training time.
• Extensive experiments demonstrated the effectiveness of both developed control strategies.

Conclusions:

• The developed sEMG-based force control strategies offer a promising alternative to traditional position control for power-assist exoskeletons.
• The individualized LDA classifier approach significantly streamlines exoskeleton control system development.
• These findings contribute to the advancement of intuitive and effective human-robot interaction in wearable robotic systems.