Iterative overlap-part estimation method for localization and generation of reliable terrain maps

For efficient wide-area exploration by an autonomous planetary rover, the generation of a global environmental map is very important. A global map is generated by matching and overlapping terrain features from digital elevation maps that are obtained during a rover traveling in a region. Each set of data contains information in different areas, and the overlapping part in each set of data may be missing data because of difference in a position and an orientation of a rover. Conventional alignment techniques based on comparing features assume that data were measured from very similar positions and orientations. Otherwise, differences in the measured region deteriorate the quality of a resulting global map. This problem should be addressed by a fairly simple way because of the limited processing capabilities of planetary exploration rovers. Hence, we propose a method for generating highly accurate global maps by considering differences in measured regions and iteratively estimating the overlapping parts in terrain data. The method is evaluated using simulated data and some experimental data obtained from experiments using a mobile robot in a real environment.