Identification of nonlinear MIMO Lagrange systems using neural networks with guaranteed persistency of excitation

Σαμάνης, Αναγνώστης; Samanis, Anagnostis

dc.contributor.author	Σαμάνης, Αναγνώστης	en
dc.contributor.author	Samanis, Anagnostis
dc.date.accessioned	2016-05-17T08:50:28Z
dc.date.available	2016-05-17T08:50:28Z
dc.date.issued	2016-05-17
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/42495
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.6743
dc.description	Εθνικό Μετσόβιο Πολυτεχνείο--Μεταπτυχιακή Εργασία. Διεπιστημονικό-Διατμηματικό Πρόγραμμα Μεταπτυχιακών Σπουδών (Δ.Π.Μ.Σ.) “Συστήματα Αυτοματισμού”	en
dc.rights	Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα	*
dc.rights.uri	http://creativecommons.org/licenses/by-nc-nd/3.0/gr/	*
dc.subject	Neural Networks	en
dc.subject	Persistency of excitation	en
dc.subject	System identification	en
dc.subject	Radial Basis Functions	en
dc.subject	Adaptive Control	en
dc.title	Identification of nonlinear MIMO Lagrange systems using neural networks with guaranteed persistency of excitation	en
heal.type	masterThesis
heal.classification	System identification	en
heal.classificationURI	http://id.loc.gov/authorities/subjects/sh85131740
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2016-02-15
heal.abstract	Motivated by learning methods in previous years and by the recent works in the control literature being done to tackle the problem of identifying nonlinear parts of unknown system dynamics, an effort is presented in this thesis to identify nonlinearities of systems, following Lagrange dynamics, in a local set of values. For this purpose, an exciting technique in a form of autonomous closed loop system is created ensuring a priori that all possible unknown dynamics of the nonlinear function of Lagrange system(to be identified) will be stimulated in a local region of values in order to be successfully learned.This autonomous system produces a a desired trajectory, which is designed (inside the local set of identification) in a way to pass through selected nodes where the centers of Radial Basis Functions (RBF) are put. Addiotionally the identification architecture is comprised of an-online Radial Basis Function (RBF) neural network identifier and an adaptive controller. The identifier estimates the unknown dynamics of the system through a processs in which, the weight estimator vector of the neural network converges to its optimal values and the learning result is stored in a linear mathematic expression. The adaptive controller is designed in a way to make sure that the system with unknown dynamics follows the desired trajectory given as input to the system, guaranteeing the satisfaction of the persistency of excitation condition for the RBF regressors employed. It is proven in this work that if the persistency of excitation condition is satisfied, then the neural network weights estimates converge to small neighborhoods of their true values,succeding in learning the actual system nonlinearities effectively. There are two identical characteristics of this approach. The first one is the fact that the autonomous system proposed produces a desired trajectory which is able to stimulate all possible state values of the system to be identified in a local region of values and the other one is the isolation between the identifier and the controller design increasing the robustness level of the learning scheme. Finally simulation examples of certain systems are provided whose nonlinear parts are identified through the identification process explained above and the effectiveness of this method is evaluated through the figures shown in each example.	en
heal.advisorName	Κυριακόπουλος, Κωνσταντίνος	en
heal.committeeMemberName	Κυριακόπουλος, Κωνσταντίνος	en
heal.committeeMemberName	Αντωνιάδης, Ιωάννης	en
heal.committeeMemberName	Παπαδόπουλος, ΕυάγγελοςPapadopoulos, Evangelos	en
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Μηχανολόγων Μηχανικών	el
heal.academicPublisherID	ntua
heal.numberOfPages	126 σ.	en
heal.fullTextAvailability	true