Wall Sensing for an Autonomous Robot With a Three-Dimensional Time-of- Flight (3-D TOF) Camera

摘要

This report describes recent work in the field of robotic perception research. The objectives of this research are to (1) program a computer to extract geometric planes characteristic of architectural walls in data from a three-dimensional time-of-flight camera, (2) track the planes through a stream of such data as the camera moved slowly through a plane-rich environment, and (3) serve plane descriptors to processes that desire such service. The programs were implemented at a high level using tools from the open-source Robot Operating System (ROS). Tests performed in simple and complex indoor environments demonstrated the ability to identify and track planes. Postprocessing was performed using a separate process within the ROS environment that receives updated planar information. Preliminary results are presented for dynamic scenes, though sensor motion estimates are necessary to properly quantify results.