This work focuses on the main problems of interactive object transfer between a human worker and an industrial robot: the recognition of the object with partial occlusion by barriers including the hand to the human worker, the evaluation of object grasping affordance, and coping with inaccessible grasping points. The proposed visual servoing system integrates different vision modules where each module encapsulates a number of visual algorithms responsible for visual servoing control in humanrobot collaboration. | Journal of Automation and Control Engineering Vol. 3, No. 4, August 2015 A Visual Servoing System for Interactive Human-Robot Object Transfer Ying Wang, Daniel Ewert, Rene Vossen, and Sabina Jeschke Institute Cluster IMA/ZLW & IfU, RWTH Aachen University, Aachen, Germany Email: {, , , }@ batches efficiently, it is desirable to combine the advantages of human adaptability with robotic exactness and efficiency. Such close cooperation has not yet been possible because of the high risk of endangerment caused by conventional industrial robots. In consequence, robot and human work areas had strictly been separated and fenced off. To enable a closer cooperation, robot manufacturers now develop lightweight robots for safe interaction. The light-weight design permits mobility at low power consumption, introduces additional mechanical compliance to the joints and applies sensor redundancy, in order to ensure the safety of humans in case of robot failure. These robots allow for seamless integration of the work areas of human workers and robots and therefore enable new ways of human-robot cooperation and interaction. Here, the vision is to have human and robot workers work side by side and collaborate as intuitively as human workers would among themselves [1]-[4]. Among all forms of human-robot cooperation, interactive object transfer is one of the most common and fundamental tasks and it is also a very complex and thus difficult one. One major problem for robotic vision systems is visual occlusion, as it dramatically lowers the chance to recognize the target out of a group of objects and then perform successive manipulations on the target. Even without any occlusion, objects in a special position and orientation or close to a human, make it difficult for the robot to find accessible grasping points. Besides, in the case of multiple available grasping points, the robot is confronted with the challenge of
