Robust nonlinear adaptive observer design using dynamic recurrent neural networks

A robust adaptive observer for a class of nonlinear systems is proposed based on a generalized dynamic recurrent neural networks (DRNN), which does not require off-line training phase. The observer stability and boundedness of the state estimates and NN weights are proven. No exact knowledge of the nonlinear matching uncertain function, such as output matching or linear-parameterized condition in the observed system, are assumed. Simulation results show the effectiveness of the proposed DRNN observer.