Terrain Classification Techniques From Ladar Data For Autonomous Navigation

Abstract

Autonomous navigation remains a considerable challenge, primarily because of the difficulty in describing the environment of the robot in a way that captures the variability of natural environments. In this paper, we focus on the problem of extracting the ground terrain surface from sparse 3-D data from LADAR mobility sensors, including the segmentation of the terrain from obscuring vegetation. In this paper, we briefly review possible approaches to LADAR processing, discuss their limitations, and describe our current approach. Results obtained with the GDRS CTA LADAR are presented