Probabilistic Terrain Mapping for Mobile Robots with Uncertain Localization

Mobile robots build on accurate, real-time mappingwith onboard range sensors to achieve autonomous navigation over rough terrain. Existing approaches often rely on absolute localization based on tracking of external geometric or visual features. To circumvent the reliability-issues of these approaches, we propose a novel terrain mapping method which bases on proprioceptive localization from kinematic and inertial measurementsonly. The proposed method incorporates the drift and uncertainties of the state estimation and a noise model of the distance sensor. Unmanned ground vehicles (UGVs) require knowledgeof the surrounding to safely and efficiently navigate through an environment. In rough terrain, an accurate terrainmodel is essential for a robot to plan a motion over and around obstacles