Recursive parameter identification of constrained systems: an application to electrically stimulated muscle

Area of Science:

• Biomedical Engineering
• Control Systems
• Signal Processing

Background:

• Real-time identification of biomedical systems often ignores known model information and physical constraints.
• Existing parameter-identification algorithms struggle to integrate constraints easily, leading to suboptimal performance.

Purpose of the Study:

• To develop a method for incorporating linear equality constraints into recursive parameter-identification algorithms.
• To enhance the accuracy of parameter estimates and improve output prediction in biomedical systems.

Main Methods:

• Developed a method applicable to most recursive parameter-identification algorithms, enforcing linear equality constraints.
• Employed orthogonal projection of parameter estimates onto the constraint surface at each time step.
• Applied the method to identify autoregressive moving-average (ARMA)-type time series models with time-varying constraints.

Main Results:

• The proposed method enhances output prediction accuracy by enforcing constraints on model parameters.
• Orthogonal projection improves parameter estimation without disrupting natural model parameterization or requiring initial simplifications.
• Demonstrated improved experimental predictive quality in identifying electrically stimulated quadriceps muscles in paraplegic subjects.

Conclusions:

• Incorporating physical constraints via orthogonal projection is an effective strategy for real-time biomedical system identification.
• This method offers advantages for predictive adaptive controllers by improving future output prediction.
• The technique successfully identified a nonlinear recruitment characteristic and muscle response dynamics in a clinical application.