Grasp Planning for Multi-Fingered Hand in Blind Grasping

Abstract

A method of blind grasping is provided in this paper, using force sensors to predict a stable robotic grasp given geometric information, but no visual information about the object being grasped. A mathematical model between fingers and the object to be grasped is constructed, based on which force closure and relative theories are used to find out an optimal planning for grasping the certain object. Firstly grasping positions are determined, in order to get positions reaching minimum force and maximum stability; afterwards, the corresponding minimum force is computed, including its magnitude and direction, using a method of exterior penalty function for optimization. Experiments indicate that this method, which requires little input, is useful in predicting a stable grasp for many shapes of objects.