Non-linear model predictive control for models with local information and uncertainties

Gaussian processes are a probabilistic, non-parametric approach to modelling that allows easy merging of ordinary measured data and local linear models. This can be of particular importance in the identification of non-linear dynamic systems from experimental data, where there is usually more data available around the equilibrium points and only sparse data is available far from them. The utility of the Gaussian process model for predictive control is investigated in this paper and is illustrated on a simple first-order vehicle dynamics.