Abstract—Autonomous driving in off-road environments requires
an exceptionally capable sensor system, particularly given
that the unstructured environment does not provide many of the
cues available in on-road environments. This paper presents a
complex vision system, which is able to provide the two basic
sensorial capabilities needed by autonomous vehicle navigation in
extreme environments: obstacle detection and path detection. A
variable-width-baseline (up to 1.5 m) single-frame stereo system
is used for pitch estimation and obstacle detection, whereas a
decision-network approach is used to detect the drivable path by a
monocular vision system. The system has been field tested on the
TerraMax vehicle, which is one of the only five vehicles to complete
the 2005 Defense Advanced Research Projects Agency (DARPA)
Grand Challenge course.