Robust Self-Localization in Industrial Environments based on 3D Ceiling Structures

Abstract- This paper introduces a new perceptual model for Monte Carlo Localization (MCL). In our approach a 3D laser scanner is used to observe the ceiling. The MCL matches ceiling structures like beams, columns, air condition and lightning installation against a world model containing line and point features. Thus the localization is not effected by clutter or any kind of dynamic obstacles on the ground level. Different experiments show that our system can be used to localize robustly in factory buildings and halls. Experimental results include a 5 day run of an autonomous fork lift truck on a logistics fair in Hannove