Fusing multiple Kinects to survey shared Human-Robot-Workspaces

Abstract

Knowledge about the static and dynamic situation is essential for joint workspaces of human and robot. Surveilling the joint area helps to realize a collaboration and co-existence of human and robot. Therefore, we propose in this paper a system that redundantly monitors the workspace to perceive obstacles within the workspace including the human worker. We use multiple, distributed range sensors (i.e. Microsoft Kinect) and de-centrally pre-process the data from the sensors. These data sets are further processed and segment and cluster unknown objects. The gained geometric information can then be used in the robot controller to allow the system to react accordingly.