A Volumetric Representation for Obstacle Detection in Vegetated Terrain

This paper presents a method for 3-D based obstacle detection on autonomous vehicles navigating in vegetated environments. At its core three different methods processing the surrounding occupancy, taken at separate stages and volumetric resolutions, are combined to a reliable and broad solution. Geometric relationships are evaluated at a coarse, yet robust, volumetric representation to form an initial assessment on obstacles. Then, a more careful evaluation takes place, at finer resolutions, to determine which obstacles are part of the scene's vegetation, thus not real obstacles. Field experiments are shown to validate the method's applicability on two different autonomous vehicles: a water surface robot and a terrestrial four-wheeled one.