Multivariable identification based MPC for closed-loop glucose regulation subject to individual variability.

Area of Science:

• Biomedical Engineering
• Control Systems Engineering
• Computational Biology

Background:

• Artificial pancreas systems require robust controllers for effective insulin infusion in diabetic therapy.
• Inter- and intra-individual variability and time delays in glucose metabolism pose significant challenges for glucose control.

Purpose of the Study:

• To develop a multivariable identification based model predictive control (mi-MPC) to overcome challenges in artificial pancreas glucose regulation.
• To directly estimate and control plasma glucose concentration (PGC) for improved diabetic therapy.

Main Methods:

• An integrated glucose-insulin model was established to describe insulin absorption, glucose-insulin interaction, and glucose transport.
• A particle filtering estimator was designed to identify individual parameters and disturbances, forming an observable glucose-insulin dynamic model.
• A mi-MPC controller was developed, embedding the identified glucose-insulin dynamic model to directly control PGC.

Main Results:

• The mi-MPC method demonstrated effective glucose regulation in 30 in-silico subjects using the UVa/Padova simulator.
• The controller achieved a mean plasma glucose concentration of 7.45 mmol/L.
• The system successfully regulated glucose without requiring meal announcements.

Conclusions:

• The developed mi-MPC method, integrating an identified glucose-insulin dynamic model, offers a promising approach for artificial pancreas systems.
• This method effectively addresses glucose variability and time delays, enhancing glucose control accuracy in diabetic therapy.
• The ability to regulate glucose without meal announcements represents a significant advancement for patient convenience and therapeutic outcomes.