Generation of Laser-Quality 2D Navigation Maps from RGB-D Sensors

The use of RGB-D cameras has become an affordable solution for robot mapping and navigation in contrast to expensive 2D laser range finders. Although these sensors provide richer information about the 3D environment, most successful mapping and navigation techniques for mobile robots have been developed considering a 2D planar environment. In this paper, we present our system for 2D navigation using RGB-D sensors. The key feature of our system is the extraction of 2D laser scans out of the 3D point cloud provided by the camera that can be later used by common mapping or localization approaches. Along with the real experiments we raise the question “how far can we go with the use of RGB-D sensors for 2D navigation?” and we analyze performance and limitations of the system compared to accurate, yet expensive, laser-based systems.