The control of the human blood sugar level is one of the most complex biomedical processes, however since it is essential for the human life the proper control of its functioning has high priority.
Several distinct models were created in this field, describing this complex physiological process in different ways. One of these approaches are the artificial pancreas models, which is applicable for insulin dependent diabetes (type 1).
The main topic of my thesis is the model-based regulation of the blood sugar level of patients having type 1 diabetes. As dynamic systems, patients have distinct dynamic parameters and could only be described with nonlinear, time-dependent models that necessitate the use of modern control theory.
In general to design a controller the states and parameters of the system together with the real output is needed. To obtain these data I designed state estimators and an online method for parameter estimation. These were implemented under MATLAB. In addition to the online estimation offline parameter estimation from previously measured data is also necessary to design virtual patients, which are to represent the real patients in the model.
In this work a grey-box identification method was used to fit the model to the measured data and construct the virtual patients.