Adaptive state observer for nonlinear MIMO systems with uncertain dynamics

This paper proposes an adaptive observer design for a class of nonlinear MIMO systems with unknown nonlinearities. Neural networks (NN) with online updating weights are utilized to estimate unknown dynamics, such that the precise system model, the Lipschitz or norm-bounded assumptions on the unknown nonlinearities are not required. By developing a novel gain design methods, some constraints used in the neural-based observer and sliding-mode observer designs, i.e., Strictly Positive Real (SPR) or matching conditions, are removed. Applicability of the presented method is verified by simulations.