Parameter identification and sensitivity analysis of a lower-limb musculoskeletal model

Area of Science:

• Biomechanics
• Robotics
• Human-Robot Interaction

Background:

• Wearable sensors are crucial for joint torque estimation in human-robot interaction.
• Existing models often require too many sensors or lack real-time output for robotic control.
• There is a need for efficient and accurate joint torque estimation methods.

Purpose of the Study:

• To develop a knee joint torque estimation model using only four electromyography (EMG) sensors.
• To propose a novel method for simplifying musculoskeletal models based on sensitivity analysis.
• To enable real-time application of joint torque estimation in robotic control.

Main Methods:

• Established a knee-joint musculoskeletal model incorporating four major muscles using advanced Hill-type components.
• Employed the genetic algorithm (GA) for parameter identification of the musculoskeletal model.
• Utilized Sobol's global sensitivity analysis to assess parameter influence and developed a sensitivity-based simplification method.

Main Results:

• The proposed musculoskeletal model achieved comparable normalized root mean square error (NRMSE) through parameter identification.
• Sensitivity analysis identified key parameters, enabling effective model simplification.
• The sensitivity-based simplification method proved effective in reducing model complexity.

Conclusions:

• A simplified knee joint torque estimation model using four EMG sensors is feasible and effective.
• Sensitivity analysis is a valuable tool for simplifying complex musculoskeletal models.
• The proposed method enhances the applicability of joint torque estimation in real-time robotic control.