Data-based Modelling for Control and Optimization

Paul Van den Hof

In this presentation a brief review will be given of developments in the systems and control field over the last decade, to construct dynamical models from experimental data particularly suited for model-based control design and optimization. This development in which the Delft group has been actively involved, has witnessed a closing of the classical gap between model construction and control design, and has paved the way for applicable methods for data-based control design, including issues of model uncertainty and controller robustness. One of the main feaures is the learning of achievable performance on the basis of controlled experiments. Achievements and current status are discussed as well as the main challenges for the near future.