Enhancing the robot avateering metaphor discreetly with an assistive agent and its effect on perception

We present a modeling approach to develop an agent that assists users discreetly in teleoperation when avateering a robot via an inexpensive motion sensor. Avateering is a powerful metaphor, and can be an effective teleoperating strategy. Avateering a Humanoid Robot (HR) with no wearable device encumberment, such as using the popular Kinect/NUI motion sensor, is also desirable and very promising. However, this control scheme makes it difficult for the slave robot to make contact and interact with objects with high accuracy due to factors such as viewpoint, individually-unique and unilateral human-side control imprecision, and lack of informative tactile feedback. Our research explores the addition of an assistive agent that arbitrates user input without disrupting the overall experience and expectation. Additionally, our agent assists with maintaining a higher level of accuracy for interaction tasks, in our case, a grasping and lifting scenario. Using theWebots robot simulator, we implemented 4 assistive agents to augment the user in avateering the Darwin-OP robot. The agent iterations are described, and results of a user study are presented. We discuss user perception towards the avateering metaphor when enhanced by the agent and also when unassisted, including perceived easiness of the task, responsiveness of the robot, and accuracy.