Synergy-Based Control of Underactuated Anthropomorphic Hands

Abstract

In this paper, a grasping control strategy for an anthropomorphic robotic hand in a synergy-based framework is designed. The goal is to achieve a human-like behavior and robustness with respect to detailed information on objects shape and size. By adapting to underactuated kinematics a well-aimed method for human grasps mapping, the synergies subspace is computed. The designed control strategy allows the hand to adapt to object contours by means of coordinated fingers motion in the synergies subspace, and it is composed of a feedforward and two corrective terms to generate fingertip reference positions. The feedforward term is obtained by choosing, among the configurations contained in the set of grasps used for synergies computation, the one that is closest to the target grasp. This term ensures a suitable preshaping of the hand in the synergies subspace on the basis of object and grasp features. On the other hand, in order to adapt the grasp to the object, a weighted sum of two corrective terms is computed using the state of the motors as the feedback of the control law. The first term is designed to generate the closure of the hand towards the object. The second term is designed to optimize a grasp quality index based on force closure property. Moreover, exploiting the motor current feedback, soft synergies are implemented realizing compliance at the contact. The experiments conducted on the underactuated SCHUNK five-finger hand demonstrate the effectiveness of the method.