Identification of the human postural control system through stochastic trajectory optimization.

Area of Science:

• Biomechanics and Control Systems Engineering
• Human Postural Control Modeling

Background:

• System identification models human postural control from experimental data.
• Deterministic methods can yield unstable controllers, limiting balance explanations and robotic applications.
• Existing eigenvalue constraints struggle with nonlinear systems and large datasets.

Purpose of the Study:

• To develop a robust system identification method for human postural control.
• To overcome limitations of deterministic approaches and eigenvalue constraints.
• To ensure stable and applicable models of the human balance system.

Main Methods:

• Utilized a stochastic system model incorporating process noise.
• Employed simultaneous trajectory optimizations across multiple noise instances for parameter identification.
• Tested stochastic and deterministic methods on linear and nonlinear controller architectures.

Main Results:

• Stochastic identification closely matched experimental data, similar to deterministic methods.
• The stochastic approach successfully avoided unstable controllers identified by deterministic models.
• Stochastic identification demonstrated superior applicability to nonlinear systems and large datasets compared to eigenvalue constraints.

Conclusions:

• Stochastic system identification provides a stable and broadly applicable method for modeling human postural control.
• This technique avoids the generation of unstable controllers, improving model reliability.
• The method offers wider potential applications than eigenvalue constraints, especially for complex systems.