Model predictive control for hybrid systems

Both academia and industry have recently directed a considerable amount of research effort on hybrid systems. Hybrid systems typically arise when continuous plants are coupled with controllers that involve discrete logic actions. Although hybrid systems are encountered in many practical situations, up to now most controllers for such systems are designed using ad hoc and heuristic procedures. Due to the complex nature of hybrid systems, it is infeasible to come up with generally applicable control design methods.

In this project we focus on structured control design methods for specific classes of hybrid systems that are industrially relevant. These methods will be extensions of the model predictive control (MPC) framework for continuous systems, so as to include hybrid systems. The MPC scheme is nowadays very popular in the oil refining and (petrochemical) process industry and has adequately proved its usefulness in practice. MPC offers attractive features that makes this control approach also interesting and relevant for extension to hybrid systems.

In this project we will develop high performance MPC controller design techniques for hybrid systems.

Currently, we have already obtained some initial results on MPC for special subclasses of hybrid systems, viz. piecewise-affine systems and max-plus linear systems. In this project we keep on extending these results to other relevant classes of hybrid systems, and we thoroughly investigate and formalize the design process, improve optimization procedures to realize real-time implementation, and use the results for practical problems of the partners from industry.

Project members: T.J.J. van den Boom (Ton), B. De Schutter (Bart)

Model predictive control, Hybrid and nonlinear systems, Model-based control, Hybrid systems