A state space I/O map to transform nonlinear systems into an observable form

Abstract

This paper deals with the problematic of state-space realisation of input-output (i/o) nonlinear systems. For that, we propose a general method that can be used to verify the realisability of i/o models and provides the equivalent state-space model when it is possible. The main advantage of the proposed approach is the use of graphic map to design the state space model without analytic analysis. The resulted state-space model is identified using a special class of modular feed-forward neural networks that embeds the state vector of the model. New results are also proposed using a realisable non-recurrent i/o subclass that is adapted to Matlab identification procedure. This approach is applied on some examples and proved its efficiency.