A computationally efficient reformulation for Data-Enabled Predictive Control
Résumé
This work investigates the computational efficiency of Data-EnablEd Predictive Control (DeePC) reformulations.
Based on Willems' fundamental lemma, this control method uses Hankel matrices to represent system dynamics.
The size---in particular the number of columns---of these Hankel matrices can incur a significant computational complexity, which has seen several attempts at being reduced.
We propose a reformulation of DeePC aiming for lower complexity and show how it allows for exponential forgetting of online collected data. The method's effectiveness is illustrated by results obtained both in simulation and experimentally.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|