Variable structure-based controllers applied to the modified Hovorka model for type 1 diabetes

Despite the fact that diabetes treatment technology is constantly evolving, research groups continue to highlight the problem of blood glucose concentration management as a critical concern. Uncontrolled blood glucose levels in the body can result from pancreas failure. The majority of diabetic patients are unable to maintain an optimal glucose concentration level and effective control is required to improve diabetic treatment. In the formulation of a control algorithm for an artificial pancreas, the modified Hovorka model was recently introduced to further explain the glucose-insulin dynamics. In this research paper, different control strategies have been implemented and compared: sliding mode control, integral and double integral sliding mode control and proportional integral derivative control used as a baseline. The goal is the evaluation of nonlinear controllers using Lyapunov theory, applied to a biologically relevant nonlinear complex model able to describe the interaction between blood glucose and insulin, based on the Hovorka equations. The proposed controllers have to achieve the desired reference level of stability and robustness in glucose concentration regulation. The key performance indexes adopted are chattering, settling time and steady-state error, also considering computational complexities. A genetic algorithm has also been adopted for parameter optimization to improve each controller's performance. The effect of perturbations in the form of food consumption was also analyzed and compared for the different control strategies.