Point Cloud-Based Spacecraft Self-Localization for Distant Small Body Exploration

Abstract

Image-based terrain-relative localization is expected to play an important role in precision navigation in various small-body missions. In the Hayabusa2 mission, accurate navigation was achieved by manual matching between the asteroid image and a set of vertices of a shape model of the target body, called a “point cloud.” The manual process intentionally excludes invisible regions and adopts visible parts for matching. In this paper, an autonomous optical localization using the point cloud, which will take the place of the manually matching process, is proposed. The proposed method performs matching with the asteroid rims derived from the captured image and point cloud, respectively. Furthermore, the proposed method allows selecting the area used for matching by adding 1-bit information, representing the visible or invisible flag, to the point cloud. The proposed method has achieved the position estimation accuracy of 0.23 [px] (1σ) in simulations with various shading conditions. This paper applies the proposed method to the Hayabusa2 flight data and shows that the proposed method bears comparison with the Hayabusa2 team estimation results.