In Silico Test for MPC and SMC Controllers Under Parametric Variations in Type 1 Diabetic Patients

JUAN E. SERENO; CAICEDO, MICHELLE A.; PABLO S. RIVADENEIRA; OSCAR E. CAMACHO

Buenos Aires

Conferencia; 2018 Argentine Conference on Automatic Control (AADECA); 2018

AADECA

Physiological parameters of glucose-insulin models for type 1 diabetes mellitus patients are assumed as time-invariant for glucose regulation control design without considering typical intra-day variations in a diabetic patient. This work analyzes the performance of two control strategies. The first one is a zone model predictive controller, the second is a sliding mode controller. Both controllers were set assuming a nominal plant model identified with standard information, continuous glucose monitoring, exogenous insulin, and carbohydrate counting. The controller's design is evaluated under variations in the most important model parameters, up to 60% in insulin sensitivity, insulin time action and carbohydrate absorption time. Results show that given nominal conditions, the predictive controller has better performance avoiding hyper and hypoglycemic events. However, under parametric variations in the model, the predictive controller is not capable of keeping its performance. Thereby, in the opposite case, the sliding mode controller achieves to maintain results for nominal conditions. This aims to study the future development of a hybrid strategy where advantages of model predictive control and sliding modes control could be taken to improve the system response for meal intake and parameter variations.