Practical Economic MPC

As opposed to tracking Model Predictive Control (MPC), economic MPC directly optimizes a given performance objective rather than penalizing the distance from a reference. On the one hand this typically improves performance. On the other hand, however, this also poses challenges in terms of stability and computational burden. In order to make economic MPC implementable in practice, we recently proposed some techniques that address both issues. In this contribution, we give a brief overview on economic MPC and the related literature and discuss how to exploit existing results in order to apply economic MPC in practice. Finally, we present a new promising research direction towards data-driven economic MPC: the use of reinforcement learning to obtain optimality for the real system, rather than for the (unavoidably inaccurate) model used by MPC.