Nonlinear MPC for Combustion Engine Control: A parameter study for realizing real-time feasibility

Thivaharan Albin, Felix Frank, Dennis Ritter, Dirk Abel, Rien Quirynen, Moritz Diehl

Abstract: Nonlinear Model Predictive Control (NMPC) schemes offer the possibility to handle complex system dynamics and advanced requirements on control, such as the consideration of constraints for a multiple input multiple output system. The major bottleneck for these algorithms is the computation time. In this paper it is shown how the different ingredients of NMPC have to be designed such that a fast NMPC algorithm can be realized which allows real-time feasibility. As the numerical example, an engine control task is investigated with high demands on the control requirements. Fast reference tracking has to be realized, while respecting constraints on the system states to prevent damages. At the same time a very nonlinear behavior is present.