Robot Motion in Unstructured Environment

To operate an autonomous vehicle in off-road terrain is one of the most difficult problems in robotic. Since it is not possible to build a high resolution model(3d map) of an unstructured terrain, in particular of an industrial area or a construction site, a robot, driving through this terrain, in particular of an industrial area or a construction site a robot, driving through this terrain, has to be able to adapt its motion to the real world. In dornier’s experimental program for “outdoor robotics applications” therefore a key function is real time obstacle deletion and avoidance. In conjunction with a gyro-system and an optical tracker this function ensures a cross country driving on a preplanned path by surveying the real world. Using an active (time of flight) laser- scanner a temporary 3d-image, consisting of 128x64 voxels, is generated four times a second and real time transformed into a 2d-obstacle-map. This map is used by the navigation-computer to avoid obstacles while following the given path as close as possible.