Optimum grasp planner and vision-guided grasping using a three-finger hand

Purpose - The capability to perform dexterous operations in an autonomous manner would greatly enhance the productivity of robotic operations. In this paper, we present a new methodology for vision-based grasping of objects or parts using a three-finger hand as a gripper of a robotic manipulator. Design/methodology/approach - The hand employed in our work, called SARAH, was designed for robotic operations on the space station, however, the main steps of our procedure can be applied for tasks in a manufacturing environment. Our methodology involves two principal stages: automatic synthesis of grasps for planar and revolute objects with SARAH and vision-based pose estimation of the object to be grasped. For both stages, we assume that a model of the object is available off-line. Findings - In the paper, numerical results are presented for grasp synthesis of several objects with SARAH to demonstrate the feasibility and optimality of the synthesized grasps. Experimental results are also obtained with SARAH as the end-effector of a seven-degree-of-freedom robotic arm, demonstrating the feasibility of the integrated vision-based grasping. Research limitations/implications - The methodology described in the paper, although represents a substantial step towards automated grasping with a robotic manipulator, still requires some decision making from the user. Further work can improve the pose identification aspects of the algorithm to make them more robust and free of human intervention. As well, the grasp synthesis procedure can be expanded to handle more complex and possibly moving objects, as well as to allow for different grasp types than those considered here. Practical implications - The work demonstrates feasibility of autonomous grasp execution in industrial setting by using a three-finger hand as a robotic gripper Originality/value - The results presented in the paper demonstrate the feasibility of synthesising optimised grasps which take into account the kinematics of the gripper. We also demonstrate a real implementation of vision-based grasping by using a robotic manipulator with a three-finger hand.