A Model Predictive Control-based Framework for Systems with State Machines

In industry, there is a persistent tendency towards more cost- and energy efficient systems. This is certainly the case for systems existing of multiple actors designed to achieve a mutual goal. Often, these actors are controlled as state machines.

Solving a central Model Predictive Control problem is a promising technique to control these systems as efficiently as possible. This technique takes into account both the models of the actors, the system limitations and the predictions for this system. Based on this information, it seeks a global optimum in real time that is applied to the actors. However, the problem for this class of systems is difficult to solve. The state machines are usually represented in the problem formulation as variables with an integer or Boolean value. This leads to a mixed-integer problem which is numerically difficult to solve.

The goal of this research project is to develop: (1) A industrially useable approach for modelling state-machines in an Model Predictive Control formulation (2) An overview of solution methods that are able to handle this type of problems.