Human-in-the-loop optimization of wearable device parameters using an EMG-based objective function.

Area of Science:

• Biomechanics
• Robotics
• Human-Computer Interaction

Background:

• Wearable robots aim to enhance user motion and experience by minimizing energetic cost (EC).
• Measuring EC typically requires indirect methods like respiratory gas analysis, which is challenging.
• Existing methods for optimizing exoskeleton assistance often rely on difficult-to-obtain physiological measurements.

Purpose of the Study:

• To introduce and validate a novel electromyography (EMG)-based objective function for estimating energetic expenditure during walking.
• To demonstrate the similarity of the EMG-based objective function to measured EC.
• To optimize wearable robot assistance using this new objective function for improved human-robot interaction.

Main Methods:

• Developed an objective function combining EMG intensity and muscle synergies.
• Validated the objective function offline by comparing it to measured EC during walking.
• Utilized an online Bayesian optimization algorithm to minimize the EMG-based objective function, optimizing walking step frequency.

Main Results:

• The proposed EMG-based objective function showed strong correlation with measured EC.
• Optimization using the new objective function required fewer iterations and reduced convergence time compared to gradient descent.
• The algorithm converged to an optimal step frequency near the user's preferred frequency, aiding EC reduction.

Conclusions:

• The novel EMG-based objective function provides a reliable and consistent estimation of energetic expenditure.
• This method offers a practical alternative to respiratory gas analysis for optimizing wearable robots.
• The approach has the potential to enhance lower limb exoskeleton assistance, improving user performance and human-robot interaction.