Quasi in-silico validations of a nonlinear LPV model-based robust glucose control algorithm for type I diabetes

Generating optimal control algorithms for artificial pancreases is an intensively researched problem. In this paper the feedback control of type 1 diabetic patients using subcutaneous insulin delivery and subcutaneous glucose monitoring is considered on the high-complexity modified nonlinear diabetic patient Sorensen-model. An acceptable compromise between the model's complexity and the developed control algorithm is certainly the choice of the parametrically varying system (LPV) description. A recently developed nonlinear model-based LPV robust controller is used and its efficiency is tested in quasi in silico mode. Min. 1 week's real data of 30 type 1 diabetic patient (aged between 6-52 years) equipped with Medtronic insulin pump were compared with simulation results of the control algorithm using static and dynamic glucose absorption profiles. The developed framework kept blood glucose level more than 90% of the time inside the 4-8 mmol/l interval (without any recalibration of the algorithm) proving its robustness. © 2011 IFAC.