Outdoor 3D map generation based on planar feature for autonomous vehicle navigation in urban environment

This paper describes a 3D textured map generation method for autonomous vehicle in urban outdoor environment, where GPS signals can not be reached. Constructed map will be used for short cycle and accurate localization and for obstacle detection using onbody laser scanner. In order to build dense 3D polygon map, planar feature of laser scanner input is extracted. They are associated and transformation matrices in between each scan point were iteratively solved. Aligned points were converted into texture mapped 3D polygons. 400×350[m] area in Univ. of Tokyo were scanned at 59 scan points, and 3D polygon map consists of 14M polygon were obtained. Experimental results of localization and autonomous path following two-wheeled inverted mobile robot PMR are shown.