Switching between multiple performance criteria to improve performance

A fundamental issue in model predictive control is the choice of the prediction and/or control horizon. Prior knowledge is often required. The transient performance is dependent on this choice. Moreover, a single choice is perhaps not the best option at all time instants. In this paper, we resolve the issue by introducing the novel idea of multiple performance criteria with switching. This method chooses automatically at every instant the best prediction and control horizon resulting in an improved transient response. The technique is then applied to the Shell benchmark control problem. We demonstrate that the tracking error is considerably reduced with lesser settling times.