Grasp Planning for Three-Fingered Robot Hands using Taxonomy-Based Preformed Grasp and Object Primitives

摘要

In this paper, we present a grasp planning method using grasp taxonomy and object primitives. Our grasp taxonomy includes newly defined grasp methods such as thumb supported pinch and palm supported pinch, to enhance grasp robustness. On the target surface, locations of finger-print that will be contacted by the robot fingers are sampled. The sampling is made to be consistent to the grasp taxonomy, called preformed grasps, matched to the target object. We perform simulations to examine the validity and the efficacy of the proposed grasp planning method.