Stable state dependent Riccati equation neural observer for a class of nonlinear systems

This paper proposes a new methodology for state estimation of a class of nonlinear systems. Owing to hard nonlinearity of some complex systems, the performance of linearisation methods is limited, thus this paper focuses on a new technique for nonlinear state estimation based on the combination of the state dependent Riccati equation (SDRE) and the neural network. SDRE technique is adapted to bring the certain nonlinear parts of system into linear-like structure and the unknown nonlinearities are estimated by artificial neural network whose weights are adjusted with guaranteed stability of the closed-loop system. This technique is strongly distinguishing the effect of uncertain nonlinearities and unmodelled dynamics to prevent divergence of state estimation error. Both the stability of the closed-loop system and uniform ultimate boundedness of the observer error are guaranteed based on Lyapunov theory.