Abstract
We propose a novel approach to construct a grasp database used on contact region information. This method is classifying grasping postures by contact regions on human hand's surface when grasping objects. Consequently, the grasping posture is generated by the information of contact regions on the hand and corresponding points on the object. It does not depend on the information of shape or size of the objects. Therefore, it can easily generate the grasping postures for various shaped objects. In this paper, we explain the structure of grasp database and the method of generating grasp postures using the database, and show experimental results of grasp synthesis for various shaped objects.
