Abstract
For an autonomous mobile robot, an important task to accomplish while maneuvering in outdoor rugged environments is terrain traversability analyzing. Due to the large variety of terrain, a general representation cannot be obtained a priori. Thus, the ability to determine the traversability based on the vehicle motion information and its environments is necessary, and more likely to enable access to interesting sites while insuring the soundness and stability of the mobile robot. We introduce a novel method which can predict motion information based on extracted image features from outdoor university campus environments, to finally estimate the traversability of terrains. A wheeled mobile robot equipped with an optical sensor and an acceleration sensor was used to conduct experiments.
