Neighbouring extremals for measurement-based optimization and predictive control

GROS Sébastien, SRINIVASAN Bala, BONVIN Dominique

An optimal solution typically consists of several arcs, with certain of them are determined by the constraints and the others by sensitivities. Using measurements to compensate for the uncertainty is straightforward for the constraint-seeking arcs, while for the sensitivity-seeking arcs this project uses the idea of neighbouring extremals. Neighbouring extremals is a rather old technique, from the 1970’s, which utilizes the local linearization of the nonlinear system. The innovative part of the project is the adaptation of this technique to singular problems and its use in the context of measurement-based optimization and robust predictive control.