Visual SLAM for Asteroid Relative Navigation

Mehregan Dor, Katherine A. Skinner, Travis Driver, Panagiotis Tsiotras; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2021, pp. 2066-2075

Abstract


This paper focuses on the application of visual SLAM for the purpose of precise autonomous navigation around an asteroid. We develop a factor graph-based approach allowing for incremental growth and fusion of sensor orientation measurements, Earth-relative inertial position measurements, as well as in-situ monocular camera imagery-based measurements, with an emphasis on the initialization step. Crucially, and in contrast to typical simulated scenarios found in the literature, we validate our approach using real imagery from NASA's DAWN mission to asteroid Vesta, along with navigation comparison data from the NASA NAIF SPICE kernels. Quantitative comparisons show impressive accuracy for a typical target characterization phase segment, both in terms of the estimated trajectory as well as in terms of the tracked estimated landmarks. Based on these results, this paper further supports the viability of autonomous SLAM-based navigation for deep-space asteroid missions.

Related Material


[pdf]
[bibtex]
@InProceedings{Dor_2021_CVPR, author = {Dor, Mehregan and Skinner, Katherine A. and Driver, Travis and Tsiotras, Panagiotis}, title = {Visual SLAM for Asteroid Relative Navigation}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops}, month = {June}, year = {2021}, pages = {2066-2075} }