Flexible and capable autonomous navigation for military missions requires hybrid road and offroad navigation. A vehicle may travel cross country where roads are not available, too exposed, or too damaged. It may also have to navigate existing roads when greater vehicle speeds are desired and exposure risk is acceptable. One critical component of the hybrid road/offroad navigation is the "Road from Offroad" problem: the ability of the vehicle to identify, navigate towards, and enter a paved roadway from an offroad position. A relatively fast (35 seconds) and reliable method developed for the Autonomous Land Vehicle (ALV) is presented. The algorithm runs on a general purpose processor and starts with a raster scan based image segmenter, followed by a rule-based region merging step. A classification scheme that identifies road areas is then applied to the largest regions created by the rule based merge. Simple descriptions of likely road areas in the image are then

Date Published: 10 March 1989
PDF: 7 pages
Proc. SPIE 1007, Mobile Robots III, (10 March 1989); doi: 10.1117/12.949082

Published in SPIE Proceedings Vol. 1007:
Mobile Robots III
William J. Wolfe, Editor(s)