Item

Abstract

In this thesis model-based control algorithms for blood glucose regulation in intensive care units, as well as during hyperinsulinaemic clamps, are designed using a nonlinear model of the glucose-insulin dynamics. The performance of the utilised model is evaluated, and several model parameters are identified offline using data sets obtained during clinical tests at the university hospital Magdeburg. In order to provide real-time applications, a nonlinear state observer for online estimation of the model states and two metabolic parameters is introduced. Subsequently, insulin and glucose output tracking controllers including the inverse nonlinear model are implemented. While the control variable of the insulin controller is insulin, the glucose controller administers glucose. Since the glucose-insulin dynamics are never known exactly, an accurate compensation of the true plant’s nonlinearity is not feasible using these controllers. Therefore, tracking controllers with two degrees of freedom, consisting of nonlinear input/output linearising feedforward control and stabilising linear state feedback control, are implemented. Simulation results using both the control structures referred to above are shown and the controller performances are discussed. Finally, the idea of hybrid blood glucose control, in particular necessary to accomplish a restriction of the control variables, is outlined.