TIME-VARYING CONTROLLERS FOR DIABETES TYPE 1

PATRICIO COLMEGNA; RICARDO S. SÁNCHEZ PEÑA

Paris

Conferencia; 6th International Conference on Advanced Technologies & Treatments for Diabetes; 2013

The problem of automatically controlling the glucose level in Insulin Dependent Diabetes Mellitus (IDDM) patients is considered. Several type of controllers have been tested in silico in the past, mostly Linear Time Invariant ones. It is clear that inter- and intra-patient uncertainty and the time-varying nature of the glucose-insulin dynamics are the main challenges in this problem. In addition, nonlinearities, measurement noise, sensor and actuator delay and saturation are also important issues to be considered. Hence a robust (model uncertainty-wise), time-varying approach seems an adequate control tool to apply. One of the authors has tested two time-varying methodologies: Linear Parameter Varying (LPV) and Unfalsified Control (UC), in [1] with Bergman's minimal model. Here, updated versions of the UC based on the parametrization of all H∞ controllers with the Q parameter update based on the Ellipsoidal algorithm is applied to more elaborated models, i.e. the UVa/Padova and the University of Cambridge's dynamics.