Grasp optimization with constraint of contact points number for a humanoid hand

Abstract

The humanoid hands grasp three-dimensional (3-D) objects is a fundamental issue in robotic grasping and dextrous manipulation. One of a difference between humanoid hands and ordinary manipulators in grasp planning is that humanoid hands has no more than five fingers, which means five contact points in maximal. However, if the same object is grasped by an ordinary manipulator, the contact points number will not be used as a grasp optimization constraint. In order to achieve the requirement of as few touch points as possible for humanoid hands, this paper propose a grasp optimization method with constraint of contact points number, and then determines and solves the optimization function according to certain methods. In this paper, we use human's action (Thumb Two-Finger) as an example. First, a mathematical model is established based on this action; Second, we use the minimum wrench as an optimization target to create a linear constraint equation by pseudo-distance method; Then transfer it into a linear unconstrained function by penalty function method. Finally, the final solution of the equation is obtained by using nelder-mead simplex(NMS) method. The effectiveness and efficiency of all methods are demonstrated by simulation.