Exploiting chaos in system identification and control

The thesis presents three main contributions: Two methods for system identification and a method for robustifying a chaotification method are developed in the thesis. The identification methods which are based on input-output data are used for the identification of linear, nonlinear and chaotic plants. Both identification methods construct nonlinear state equations for systems to be identified by using white Gaussian noise or chaotic state sequences in learning state recursions. Nonlinear state equations and output equations in these identification methods are realized by artificial neural networks. Extensive computer experiments on a set of benchmark plants for noisy input and initial conditons show that the developed methods provide significantly better performances when compared to the known identification methods. The third contribution of the thesis is another application in the direction of exploiting chaos now for forcing a nonchaotic plant to chaos in a robust way.