3-D incremental modeling and robot localization in a structured environment using a laser range finder

A polyhedral representation of the world is used to deal with autonomous navigation in structured environments. A description is given of algorithms needed to build snapshot models with planar faces from noisy and sparse range data, and to perform 3-D data fusion between these snapshot models, in order to build incrementally a reliable 3-D model from which a path planner could generate safe trajectories. A full implemented modeling strategy based on range data processing is presented, and the different modeling services required for indoor scene navigation by the robot HILARE-2 are described