Object-Independent Human-to-Robot Handovers Using Real Time Robotic Vision

We present an approach for safe, and object-independent human-to-robot handovers using real time robotic vision, manipulation. aim general applicability with a generic object detector, fast grasp selection algorithm, by single gripper-mounted RGB-D camera, hence not relying on external sensors. The robot is controlled via visual servoing towards the of interest. Putting high emphasis safety, we use two perception modules: human body part segmentation, hand/finger segmentation. Pixels that are deemed to belong filtered out from candidate poses, ensuring safely picks without colliding partner. selection, modules run concurrently in real-time, which allows monitoring progress. In experiments 13 objects, was able successfully take 81.9% trials.