Exploring Non-linear Dynamical Structure for Knee Kinematics Using Machine Learning

Area of Science:

• Biomechanics
• Movement Science
• Nonlinear Dynamics

Background:

• Human gait is a complex, cyclic process requiring multi-limb coordination.
• The nonlinear dynamics of knee movement during walking cannot be fully explained by linear models.

Purpose of the Study:

• To apply advanced Machine Learning (ML) techniques to uncover the governing equations of knee movement during walking.
• To utilize the Sparse Identification of Nonlinear Dynamics (SINDy) algorithm for this purpose.

Main Methods:

• Gathered single-subject knee motion data using infrared markers during normal walking.
• Employed the PySINDy library in Python to implement the SINDy algorithm.
• Determined the governing equations and calculated dynamical system coefficients for knee kinematics.

Main Results:

• The SINDy algorithm effectively identified nonlinear dynamic systems governing knee kinematics during gait.
• Governing equations that accurately describe dynamic systems in human walking were revealed.

Conclusions:

• The SINDy algorithm is a powerful tool for uncovering nonlinear dynamics in movement science.
• This approach provides new insights into the complex mechanics of human gait.