Optimal PI-fuzzy logic controller of glucose concentration using genetic algorithm

Abstract

In this paper, an optimal PI-fuzzy controller to regulate plasma glucose in Type1 diabetic patients is introduced. This controller is designed to mimic the functionality of β-cell in pancreas. Complete lack of insulin resulting from β-cell deficiency leads to a high blood glucose concentration or the so-called Type 1 diabetes. Patients having this disease need external insulin treatment to keep their blood glucose within normal ranges and to protect themselves from hyperglycemia risk. A miniaturized insulin infusion pump integrated with a continuous glucose sensor and driven by a closed-loop control algorithm can be implemented to create an artificial β-cell. For simulation purpose, the control algorithm needs a mathematical model representing the natural interaction between insulin and glucose. The up-to-date nonlinear delay differential model of glucose-insulin regulatory system, which represents the glucose-insulin metabolic system within the human body, is used as a reference and as a patient model. The controller parameters, which include membership functions and scaling parameters, are optimized by the genetic algorithm. Controller performance is evaluated thorough simulation studies and compared to that of the reference model. The results show that the plasma glucose and insulin ranges, average glucose value, and total amount of delivered insulin under the controller are very close to that of the reference model.