Interval Observers for a Class of Nonlinear Systems Using Gaussian Process Models

An interval observer design approach for partially unknown nonlinear systems is developed, where the unknown system component is modeled using Gaussian processes and noisy system measurements. The proposed method is applicable for bounded nonlinear systems where the system uncertainty is described by a Lipschitz continuous function. The interval observer generates a correct estimation error with high probability, and the error bound is decreased by employing new training data points.