Internal Model Control Based on a Gaussian Process Prior Model

To improve transparency and reduce the curse of dimensionality of non-linear black-box models, the local modelling approach was proposed. Poor transient response of Local Model Networks led to the use of non-parametrical probabilistic models such as the Gaussian Process prior approach. Recently, Gaussian Process models were applied for Minimum Variance Control. This paper introduces the use of the Gaussian Process model for non-linear Internal Model control. The invertibility of the Gaussian Process model is discussed and the use of predicted variance is illustrated on a simulated example.