Combined Stochastic and Deterministic Interval Predictor∗

The aim of this work is to propose a new model-free interval predictor for nonlinear systems. A model-free interval predictor is a method that provides an outer estimation of the future system output using stored past information of the system. The predictor does not use a parametric model to obtain the prediction. Each time instant, a linear combination of stored past outputs and an error bound are used to obtain the interval prediction. The novelty of this work is to assume a combined deterministic and stochastic assumption on the error term to obtain the interval prediction. In order to bound the error, it is assumed that the nonlinear system can be approximated locally by an unknown affine model. The center of the interval prediction can be used as point or nominal prediction. A real world example is provided to illustrate the improvement provided by the proposed predictor.