Combined stochastic and deterministic interval predictor for time-varying systems

This work proposes a new interval predictor for time-varying linear systems. An interval predictor is a method that provides an interval as outer estimation of the future system output. The center of the interval prediction can be used as point or nominal prediction. This interval center is obtained by a linear combination of stored past outputs. The interval width is obtained using an outer bound of the prediction error. Two different approaches have been considered in literature, based on deterministic and stochastic assumptions respectively. The novelty of this work is to use a combined deterministic and stochastic assumption on this bound to obtain the interval prediction. The aim is to achieve a low error in the central prediction and a small interval width. An example is provided to illustrate the improvement provided by the proposed predictor.