Robot-Human Hand-Overs in Non-Anthropomorphic Robots Prasanna Kumar Sivakumar, Chittaranjan S Srinivas School of Mechanical Engineering SASTRA University, Tanjore, India Email: prasannakumar.0091@gmail.com, chitt@live.in Andrey Kiselev and Amy Loutfi Centre for Applied Autonomous Sensor Systems ¨ Orebro University, ¨ Orebro, Sweden Email: andrey.kiselev@oru.se, amy.loutfi@oru.se Abstract—Robots that assist and interact with humans will inevitably require to successfully achieve the task of handing over objects. Whether it is to deliver desired objects for the elderly living in their homes or hand tools to a worker in a factory, the process of robot hand-overs is one worthy study within the human robot interaction community. While the study of object hand-overs have been studied in previous works [1], these works have mainly considered anthropomorphic robots, that is, robots that appear and move similar to humans. However, recent trends within robotics, and in particular domestic robotics have witnessed an increase in non-anthropomorphic robotic platforms such as moving tables [2], teleconferencing robots [3] and vacuum cleaners. The study of robot handover for non-anthropomorphic robots and in particular the study of what constitute a successful handover is at focus in this paper. For the purpose of investigation, the TurtleBot 1 , which is a moving table like device is used in a home environment. I. APPROACH The robotic platform used in the hand-overs is a TurtleBot and ressembles a cylindrical table without arms or manipu- lators.A hand-over as a whole comprises of four parts. (1) Approach – The robot from its initial point moves to a point that is away from the field of view of the user. (2) Reach – The robot from that point moves to a point near the user in a particular fashion. Each hand-over method has its own reach. (3) Cue – The robot upon reaching the user performs an intuitive cue which serves an alert to the user to pick up the object. This intuitive cue is optional for a hand-over. (4) Depart – The robot upon execution of cue waits for the user to pick-up the object for a fixed amount of time (3 seconds). Invariable of the object being picked up or not, it moves back to the point from where it started its approach. Contrasts are used to design the hand-overs where in particular we make us of two contrasts inspired by [1]. (1) Spatial Contrast – refers to the distinctness in the way the robot moves with the object on top of it during the hand-over process compared to the way it would move with the object during other actions. A reach motion with high spatial contrast is a distinct motion that conveys the intent of handing over. (2) Temporal Contrast – refers to change from one pose to another as an indication to the user to pick-up the object. Since the TurtleBot does not have arms it cannot show the change in pose while moving towards the user. It shows the change in pose only after arriving at a point near the user. A cue with 1 www.willowgarage.com Fig. 1. Left: The Shake cue during the Shake handover. The 30 degree spin, Temporal Contrast. Right: Circle hand-over with spatial contrast. high temporal contrast is a distinct act that conveys the timing of handing over. We construct three hand-over methods considering spatial and temporal contrasts. (1) A Naive hand-over where the robot starts from the initial point and approaches does the reach in a straight line to move to point B near the human, see Fig. 1. It does not execute an intuitive cue and waits for three seconds. After the wait time, invariable of the object being picked up or not, the robot leaves back to the place from where it started its approach. This method has neither a spatial nor a temporal contrast. (2) The Shake hand-over, the robot executes the shake cue. It spins to a pose approximately 15 ◦ to the left followed by a spin of 30 ◦ to the right followed by another 15 ◦ spin to the left to reach the original pose and waits for three seconds. The depart is similar to that of the Naive hand-over method. Fig. 1 Left shows the temporal contrast exhibited by the robot. Shown in green lines are the two distinct poses that are non-handing over poses and the robot at the end of the cue would reach a pose in between the green lines and that is the handing over pose. (3) The Circle hand-over moves in a circular trajectory to point B. By doing so the robot moves through the human’s front view. It does not execute a cue and waits for three seconds followed by the depart. This method has a distinct reach and has high spatial contrast and no temporal contrast. Fig. 1 right shows the spatial contrast exhibited by the Circle Handover method in the form of Circular trajectory motion which is not found in the Naive handover method. II. THE EXPERIMENT Twenty-four participants performed hand-overs with the robot. Tab. I gives a detailed account of the test participants.