Neuro tracking control for glucose-insulin interaction model

In this paper a neural-tracking control algorithm based on the synthesis of adaptive control functions which are derived to follow the dynamics of a reference model is shown. The states used in the control algorithm are given by a neural network identifier which depends on the factors: the system states dynamic, the dynamics of the identifier, a special weight learning law based on a Lyapunov stability analysis. This technique was applied to the glucose-insulin interaction model (Bergman), where two external inputs, levels of glucose concentration, insulin concentration and finally insulin concentration in the remote compartment are corresponding considered as inputs and states. The reference model was design using a small variation of the patient´s normal glucose and insulin concentrations. The algorithm efficiency is tested by numerical calculation on the system using the convergence portrait for each state time evolution and the convergence to zero of the performance index tool.