A Backstepping-Based Nonlinear Controller for Glucose-Insulin System Dynamics in Type-1 Diabetes Patients

Abstract

This paper investigates the function of the artificial pancreas, which is devised based on a dynamical backstepping approach. The Bergman's minimal model, used to describe the glucose-insulin system, has been extended to encompass the dynamics of the insulin pump and external disturbances to closely simulate real-world scenarios. Three techniques, namely feedback linearization, conventional backstepping, and super-twisting sliding-mode control, are evaluated in comparison to dynamical backstepping in the context of regulating blood glucose levels in individuals with type-1 diabetes. In order to enhance the comparison of the controllers, we have taken into account the measurement noise and faults in the insulin pump as well. Additionally, Monte-Carlo analysis is utilized as a practical tool to experimentally evaluate the robustness of the nonlinear controllers against measurement errors and variations in model parameters for different individuals, as would be encountered in a clinical trial. The extensive numerical simulations confirm that the dynamical backstepping method closely emulates the functionality of the natural pancreas and surpasses the super-twisting sliding-mode control method, the feedback linearization method, and the conventional backstepping method when faced with measurement noise, insulin pump faults, and parameter variations.