Evaluation and selection of grasp quality criteria for dexterous manipulation - ANR - Agence nationale de la recherche Accéder directement au contenu
Article Dans Une Revue Journal of Intelligent and Robotic Systems Année : 2022

Evaluation and selection of grasp quality criteria for dexterous manipulation

Résumé

The development of algorithms capable of automatically generating optimal grasp involves first of all the necessity to define the notion of optimal grasp in relation to the target task. To address this problem, the scientific community offers many quality criteria in the literature and continues to propose new ones for grasp synthesis purpose. This paper aims at proposing a synthesis and a fine analysis of the quality criteria useful to evaluate a grasp in a context of adaptive grasping, as well as in the perspective of in-hand manipulation. These criteria are divided in two categories, the first one has 11 criteria and focuses exclusively on the location of contact points while the second one has 5 criteria and takes into account the kinematics of the robotic hand as well. Evaluation of the criteria is proposed with a common evaluation framework based on reference objects and reference manipulation tasks. The evaluation and illustration of the resulting grasps with the different criteria allow to appreciate the physical meaning of each of these criteria with this common evaluation framework. In order to reduce the number of criteria to be used in the context of a grasping synthesis process, a correlation study is carried out. The results show that several criteria in the literature are strongly correlated. Four criteria are finally chosen. Thus, to demonstrate the relevance of the selected criteria, a grasp synthesis process is used for in-hand manipulation purpose. An evolutionary approach is used to solve this multi-criteria optimization problem. The approach is validated in the OpenRAVE simulation environment and then demonstrated with the new RoBioSS hand: a fully actuated dexterous robotic hand with four fingers and sixteen degrees of freedom. Experimental results illustrate the relevance of the choice of these criteria to produce robust grasps leading to stable in-hand manipulations with large amplitudes.
Fichier non déposé

Dates et versions

hal-03826741 , version 1 (24-10-2022)

Identifiants

Citer

H. Mnyussiwalla, P. Seguin, P. Vulliez, J. Gazeau. Evaluation and selection of grasp quality criteria for dexterous manipulation. Journal of Intelligent and Robotic Systems, 2022, 104 (2), pp.20. ⟨10.1007/s10846-021-01554-4⟩. ⟨hal-03826741⟩
16 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More