Vision-Based Absolute Navigation for Descent and Landing

摘要

Future space exploration missions such as Mars sample return or the lunar lander mission require precise information about the lander position during the descent and landing steps. This article presents a solution, Landstel, that exploits on-board vision to localize the lander with respect to orbital imagery. The algorithm is based on the geometric repartition of surface landmarks to match descent images with orbital images. It is designed to be robust to illumination variations, to be independent of the presence of specific features on the surface, and to have low memory and processing power requirements. Overall absolute localization is achieved by the integration of Landstel with inertial navigation or with visual odometry, according to a loose fusion scheme. Extensive results with simulated and real data validate the proposed approach, and show that it works well even under difficult conditions where the landmark repeatability rate drops to 10%.