Joint torques estimation in human gait based on Gaussian process.

Area of Science:

• Biomechanics
• Robotics
• Machine Learning

Background:

• Human gait relies on complex nervous and musculoskeletal dynamics to continuously adapt joint torques.
• Understanding these dynamics is crucial for developing advanced assistive devices and robots.

Purpose of the Study:

• To estimate lower limb joint torques during human gait.
• To utilize Gaussian processes for accurate torque prediction.

Main Methods:

• A Gaussian process (GP) based data fusion algorithm was developed.
• Joint angles served as the primary input for the GP model.
• The model explores statistical correlations between joint angles and torques.

Main Results:

• The proposed GP model accurately estimates joint torques.
• The statistical nature of the model captures underlying joint angle-torque relationships.
• Experiments validated the model across various walking speeds (0.8, 1.2, 1.6 m/s) and 5 subjects.

Conclusions:

• Gaussian processes provide a viable method for estimating human gait joint torques.
• Accurate joint torque estimation is achievable across different walking scenarios.
• This technique has potential applications in exoskeleton optimization, active prostheses control, and humanoid robotics.