Foliage Discrimination Using a Rotating Ladar

An outdoor environment presents t o a robot objects that are drivable, like tall grass and small bushes, or non-drivable, like trees and rocks. Baditionally, because of the dificulty of discriminating between these classes, a robot searches for paths free of objects of either class. Although this approach prevents collisions with objects missclassified as drivable, it also eliminates a large number of non-free drivable paths and by doing so, it might eliminate the only path to a desired destination. In this paper we present of a real time algorithm that detects foliage using range estimates from a rotating laser. Objects not classified as foliage are conservatively labeled as non-drivable obstacles. In contrast to related work that uses range statistics to classih the objects, we exploit the expected localities of an obstacle, in both space and time. The Urbie robot is presently using this algorithm to discriminate drivable grass from obstacles during outdoors autonomous navigation tasks.