Key points are not available for this paper at this time.
Robot navigation in poorly structured and uneven outdoor environments is an unsolved problem. Thus we present a SLAM (simultaneous localization and mapping) approach that is based on "leveled range scans ". The combines 3D perception with 2D localization and mapping. In this way established path planning and 2D navigation algorithms can be used in uneven terrain without the computational costs of full three dimensional modeling. The paper describes the processing steps data acquisition, obstacle segmentation, generation of leveled range scans and SLAM with these scans. Additionally, the paper shows experimental results in man-made outdoor environments as they are typical for service robots used by civilians.
Brenneke et al. (Thu,) studied this question.