Practical comparison of MPC Toolboxes

Abstract: Model predictive control can be used to control a range of processes, from selfdriving cars to chemical plants. The education of the engineering students that in the future will design these controllers is an important matter. In this thesis, three Matlab toolboxes have been evaluated and compared from a student’s perspective. The toolboxes are Mathwork’s own toolbox called Model Predictive Control Toolbox, Multi-Parametric Toolbox 3, and MATMPC. With a practical approach, three types of controllers have been created in each toolbox, a basic MPC controller, an MPC controller with integral action, and an MPC controller with gain scheduling. The documentation has been explored and the notation has been compared with the theory that is taught to the students. Different ways to implement integral action and gain scheduling have been evaluated and the controllers have been run with a real process to emulate a laboratory session and the results from the different toolboxes have been compared. The notation in Multi-Parametric Toolbox 3 did best correspond to the students’ knowledge about MPC and had in addition the best performance of the three toolboxes.