Grasp database generator for anthropomorphic robotic hands

Abstract

Grasp databases can be useful for data-driven grasp synthesis algorithm or benchmarking grasp quality metrics. This paper presents an algorithm to produce a grasp database for anthropomorphic robotic hands. The method can generate different types of grasps from the human grasp taxonomy, it relies on the notion of hand preshapes. For each type of grasp there is a hand preshape defined, containing information on how the hand should approach the object and close its fingers. The proposed approach is validated on the anthropomorphic RoBioSS hand.