An automatic assembly parts detection and grasping system for industrial manufacturing

Abstract

In this work, supported by the AUTORECON European Project, a highly reconfigurable gripper equipped with vision system for factory automation is described. The gripper has been designed to be dexterous and able to adapt itself to grasp parts within a wide range of shapes and weights. Following a similar strategy, the vision system does not need any database or training. It can be used to manage the grasping of objects that are unknown in advance, regardless from their position or alignment within the working range of the gripper. Before grasping the parts, the vision system automatically detects the position of the object. The optimal grasping points are then deduced by analyzing the geometry features of the part and the actual position and configuration of the fingers of the gripper. The detection and decision on proper grasping points is fast in the order of tens ms, so the whole system can be successfully used even if the parts are swiftly moving on a conveyor.