Towards a reactive grasping system for an industrial robot arm

Abstract

As robots are being increasingly called to operate out of well-structured environments, our contribution focuses in the grasp determination in unknown objects, vision being the exclusive mode of perception. Two strategies based on visual parameters extracted from 2D scenes are presented. The first one seeks the object's contour for a grasp that is optimal within a gravitational field, while the second one aims at locating all the grasps that meet certain contact stability criteria, ignoring the gravitational field assumption. Both strategies are complementary in the sense that they endow a robot with different manipulation skills in complex, non-structured scenarios, and can be easily integrated in the implementation of a reactive behavior for the robot.