RBF Neural Network Controller based on OLSSVR


Ucak K., Öke Günel G.

9th Asian Control Conference (ASCC), İstanbul, Turkey, 23 - 26 June 2013 identifier

  • Publication Type: Conference Paper / Full Text
  • City: İstanbul
  • Country: Turkey

Abstract

In this paper, a predictive adaptation method based on Online Least Square Support Vector Regression (OLSVR) for a RBF controller has been proposed. System Jacobian is approximated via Online LSSVR model of the system to tune RBF controller. The parameters of the controller have been tuned depending on K-step ahead future behavior of the system to provide adaptation ability to the controller under changing conditions. Levenberg Marquard algorithm is utilized as learning algorithm for controller parameters. The proposed method has been evaluated by simulations carried out on a magnetic levitation system, and the results show that the control performance has been improved.