State estimations for application in diabetes control

The aim of this work is the evaluation of two state-estimators in a glucose estimation in a diabetes control setting. Technological solutions leading to the design of an artificial pancreas heavily rely on the availability of glucose subcutaneously. Practical solutions for these measurements often suffer from significant time-delays and direct measurements of the required states is often impossible. This complicates the prediction of the required insulin dose, leading to fixed glucose intake patterns for the patient, i.e., a fixed diet at predefined times during the day. This paper describes a state-estimation approach to deal with the large time-delays, which enables relaxations of the restrictions on the patients glucose intake. Based on two models, widely used within the medical society, multiple Kalman-based filtering techniques are compared.