Robust neural adaptive observer for MIMO nonlinear systems

A robust neural adaptive observer for multivariable nonlinear dynamical systems that present an unknown general state equation and known linear output equation is proposed. A robust learning algorithm based on modification and linearly parameterized neural networks are used in order to guarantee stability of the state estimation and NN weight errors. This proposed design methodology expands the class of continuous nonlinear systems to which neural adaptive observer can be applied and the usual strictly positive real assumption is avoided. Simulations for illustrating the performance of the proposed observer are presented.