Title: A novel terminal sliding mode observer with RBF neural network for a class of nonlinear systems

Authors: Amin Sharafian; Reza Ghasemi

Addresses: Department of Electrical Engineering, University of Qom, Qom, Iran ' Department of Electrical Engineering, University of Qom, Qom, Iran

Abstract: A novel scheme for designing a new observer with combining radial basis function neural network (RBFNN) and terminal sliding mode approaches is presented. Terminal sliding mode adopted to cover the effects of internal disturbances of the system and neural network handles the problem of uncertainties and unmodelled dynamics. Convergence of the observer error to zero and accurate estimation of uncertainties of the nonlinear system are the main advantages of the proposed observer. This observer is designed based on output injection method in which the error is injected to the next state in every step until it reaches the last state of the system. Eventually, the error is suppressed and converged to zero in the last state by applying RBFNN. The stability of neural network weights which are updated adaptively and the error dynamic are guaranteed by the Lyapunov theory. Finally, the simulation result shows the promising performances of the proposed observer.

Keywords: RBFNN; state estimation; neural observer; nonlinear systems; terminal sliding mode observer; output injection.

DOI: 10.1504/IJSCC.2018.095269

International Journal of Systems, Control and Communications, 2018 Vol.9 No.4, pp.369 - 385

Received: 13 Mar 2017
Accepted: 25 Mar 2018

Published online: 16 Aug 2018 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article