Laser-vision cooperation for map building and landmarks recognition

Abstract

Describes a map-building and landmarks-recognition method in an autonomous vehicle's environment for the purpose of navigation and localization. In order to perform sensor-based navigation, simple and real time map building is needed, while for mobile robot localization, a more complete representation of the environment showing landmarks is necessary. Within this vision, a cooperating sensor system is proposed. A local map composed of segment and corner primitives is first generated in real time from 2D laser scanner sensory information. A vision system then verifies particular corner primitives and identifies them as obstacles or as valid landmarks used by the localization task. This system is used by the mobile robot ROMANE, on which the real experimentation is carried out.