An RBFN-based observer for nonlinear systems via deterministic learning

Recently, it was shown that for a class of nonlinear systems with only output measurements, by using a high-gain observer and a dynamical radial basis function network (RBFN), locally-accurate identification of the underlying system dynamics can be achieved along the estimated state trajectory. In this paper, it will be shown that the learned knowledge on system dynamics can be reused in an RBFN-based nonlinear observer, so that correct state estimation can be achieved not by using high gain domination, but by the internal matching of the underlying system dynamics. The significance of the paper is that it shows that non-high-gain state estimation can be achieved by incorporating the knowledge reuse mechanism of the deterministic learning theory. Simulation studies are included to demonstrate the effectiveness of the approach.