Automatic Artificial Pancreas Systems Using an Intelligent Multiple-Model PID Strategy

Area of Science:

• Biomedical Engineering
• Control Systems Engineering
• Endocrinology

Background:

• Type 1 diabetes requires continuous glucose monitoring and insulin delivery.
• Artificial pancreas (AP) systems aim to automate glycemic regulation.
• Existing AP systems face challenges in personalized control and safety.

Purpose of the Study:

• To propose an individualized intelligent multiple-model technique for designing automatic artificial pancreas (AP) systems.
• To enhance glycemic regulation in type 1 diabetic patients.
• To improve the safety and performance of AP systems.

Main Methods:

• Mathematical identification of the insulin-glucose regulatory system using multiple local models.
• Bi-objective optimization to balance model number and system complexity.
• Genetic algorithm (GA) for tuning proportional-integral-derivative (PID) controllers.
• Fuzzy gain scheduling for controller participation and safety mechanisms (IOB constraint, pump shut-off).

Main Results:

• The proposed methodology successfully designed AP systems.
• In silico experiments demonstrated limited hyperglycemia.
• The system effectively prevented severe hypoglycemia in virtual patients.

Conclusions:

• The intelligent multiple-model technique offers a robust approach for AP system design.
• The developed AP systems show significant potential for improving glycemic control in type 1 diabetes.
• The integration of safety mechanisms enhances AP system reliability and patient safety.