Predictive algorithms for neuromuscular control of human locomotion

Area of Science:

• Biomechanics
• Computational Science
• Human Movement Analysis

Background:

• Quantifying human muscular activity is crucial for biomechanical analysis.
• Current methods for large-scale systems are computationally expensive and slow.
• Non-derivative techniques require numerous integrations, limiting practicality.

Purpose of the Study:

• To develop an efficient numerical algorithm for biomechanical optimal control problems.
• To improve the speed and reduce the computational cost of quantifying muscular activity.
• To enable practical analysis of large-scale biomechanical systems.

Main Methods:

• Applied direct collocation with trapezoidal discretization to convert equations of motion into algebraic constraints.
• Utilized an augmented Lagrangian formulation for optimization with equality and inequality constraints.
• Solved the min-max problem using a generalized Newton method with analytical first- and second-derivative information for local quadratic convergence.

Main Results:

• Demonstrated efficacy on a steady-state pedaling problem with 7 segments and 18 muscle groups.
• Computed muscle activations showed good agreement with experimental electromyography (EMG) data.
• Achieved significant reduction in computational effort and solution times.

Conclusions:

• The proposed numerical algorithm is efficient and accurate for biomechanical optimal control.
• The method offers a practical solution for analyzing large-scale human movement.
• Analytical derivatives improve convergence speed and reduce computational burden.