Grasp Transfer for Deformable Objects by Functional Map Correspondence
Résumé
Handling object deformations for robotic grasping is still a major problem to solve. In this paper, we propose an efficient learning-free solution for adapting grasp hypothesis to deformed versions of an object. To this end, we investigate the applicability of functional map (FM) correspondence, where the shape matching problem is treated as searching for correspondences between geometric functions in a reduced basis. For a user selected region of an object, we apply the local contact moment (LoCoMo) grasp planner to generate grasp candidates. Next, these candidates are transferred to an instance of the object that has suffered an arbitrary level of deformation. Finally, the best feasible grasp, is executed on the object while respecting the original finger configuration as much as possible. Experimental validation in simulation shows the efficiency in our approach.
Domaines
Automatique / RobotiqueOrigine | Fichiers produits par l'(les) auteur(s) |
---|