Generalized extended state observer design for the estimation of the rate of glucose appearance in artificial pancreas

Postprandial control is a relevant challenge in the context of glycemia regulation with an artificial pancreas for patients with type 1 diabetes. In general, pre-meal bolus schemes are adopted to cope with this problem. These schemes usually require individuals to announce the meal to the controller. As a result, control performance will be degraded if meal announcement is skipped or carbohydrates are misestimated by patients. An estimation of the rate of appearance could improve postprandial control without depending on meal announcement, since this variable could be employed either in meal detection algorithms, or in feedforward disturbance compensation techniques. This paper considers the viability of Generalized Extended State Observer in the estimation of the rate of appearance. A tuning procedure is presented to avoid overshoots in the estimation. Lastly, the observer is evaluated in an in-silico validation carried out by UVa-Padova simulator.