Algorithmes de planification de trajectoires robustes pour des tâches robotiques en présence d'incertitudes paramétriques
le 27 avril 2023
- IRISA / Inria de Rennes en salle Turing-Petri
Soutenance de thèse de Pascal Brault (Université de Rennes / Laboratoire IRISA), ancien élève (promotion 2015), ATER au département Mécatronique
Spécialité : Automatique, signal, productique, robotique
Abstract:
One of the major challenges of automated systems is the need to operate under real-world, thus uncertain conditions. Parametric uncertainty is a common problem, and arises when performing various robotic tasks. In this thesis, we explore the possibilities provided by the planning of trajectories, whose tracking is inherently robust to uncertainties. In the first contribution, we extend minimum sensitivity trajectory planning algorithms by introducing the input sensitivity, which, when reduced, leads to an increased prediction of the system input. This problem, although treated for aerial robotics, is generalised for any system. As the sensitivity problem relies on the assumption of perfect knowledge of the state, in the second contribution, we combine the control-aware and observability-aware algorithms thanks to the Chebyshev method, in order to generate trajectories robust to uncertainties, that also ensure a better estimation of the system's variables/parameters. Finally, in the last contribution, we develop a theory that exploits the parametric uncertainty data, and construct the worst case deviation tubes around the nominal system's variables. This new expression of the problem increases the reliability of systems, as we ensure feasibility even for the largest parametric deviations.
> Voir la soutenance en vidéo
- Thématique(s)
- Recherche - Valorisation
- Contact
- Pascal Brault
Mise à jour le 3 mai 2023
Jury
- Isabelle FANTONI (Rapporteuse), Directrice de recherche CNRS, LS2N, Nantes
- Cristian SECCHI (Rapporteur), Full professor in robotics, University of Modena and Reggio Emilia
- Fabio MORBIDI (Examinateur), Maître de conférences, Université Picardie Jules Verne, Amiens
- François CHAUMETTE (Examinateur), Directeur de recherche Inria, Rennes
- Paolo ROBUFFO GIORDANO (Directeur), Directeur de recherche CNRS, IRISA, Rennes