SLAM for Autonomous Planetary Rovers with Global Localization

Abstract

This paper describes a novel approach to SLAM techniques applied to the autonomous planetary rover exploration scenario in order to reduce both the relative and absolute localization errors, using two well proven techniques: particle filters and scan matching. Continuous relative localization is improved by matching high resolution sensor scans to the online created local map. Additionally, in order to avoid issues with drifting localization, absolute localization is globally corrected at discrete times, according to predefined event criteria, by matching the current local map to the orbiter's global map. The resolutions of local and global maps can be appropriately chosen for computation and accuracy purposes. Further, the online generated local map, of the form of a structured elevation grid map, can also be used to evaluate the traversability of the surrounding environment and allow for continuous navigation. The objective of this work is to support long-range low-supervision planetary exploration. The implemented SLAM technique has been validated with a dataset acquired during a field test campaign performed at the Teide Volcano on the island of Tenerife, representative of a Mars/Moon exploration scenario.