Grasp Planning for Multi-Fingered Hands Based on Rectangular Convex Models(Mechanical Systems)

Abstract

This paper discusses the grasp planning for a multifingered hand attached at the tip of a humanoid robot's arm. Different from previous methods on grasp planning, we use the reculangular convex models for both the object and the hand grasping region. By using these models, we provide heuristic methods for determining both a grasping style and nominal position/orientation of the palm. With this method, we can easily obtain the nominal position/orientation of the palm. The final grasping posture can be obtained by testing the force closure just around the nominal position/orientation of the palm. We will show the effectiveness of the proposed method by using several numerical examples. Also, by using the four-fingered hand and the stereo vision system both equipped with the humanoid robot HRP-3P, we will perform the experiment of grasping an object.