Modeling, identification and state estimation for artificial pancreas. Part II: Study of the estimation problem.

GODOY, J.L.; P. ABUIN; P. S. RIVADENEIRA; A. H. GONZÁLEZ

Buenos Aires

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

AADECA

Type 1 Diabetes is currently controlled by many model-based advanced control techniques, embedded in what is known as artificial pancreas (AP). In this context, the problem of the identification and state estimation is crucial to account for the effect of measured, partially-measured and unmeasuredinputs/disturbances. In this work, a diabetic patient minimal model with long-term validity is proposed and, based on this model, several identifications are performed. Furthermore, and no less important, different estimators are designed to obtain the states in several scenarios, including unannounced meals andmeasurements errors in the glycemia and insulin injections. The properties/benefits of the modeling, identification and estimation are discussed in detail in this two-parts work: Part I describes the model and the identification, while Part II is devoted to study the estimation problem. Both parts include several simulation results - based on real and virtual patient data - which demonstrate the pros and cons of the proposed strategies.