Autonomous Vision-Based Navigation at Small Bodies Combining Centroiding and Visual Odometry

Vision-based navigation is a key technology for spacecraft operating in challenging conditions, such as those encountered at small bodies. To date, most navigation tasks are carried out on the ground, under the supervision of expert operators. However, onboard automation of routine tasks reduces mission costs and at the same time increases the scientific return by avoiding communication delays. In this work, an autonomous vision-based navigation architecture is proposed that exploits two complementary measurements: centroiding and visual odometry. When combined, they provide both absolute and relative navigation information, which can be used to navigate the spacecraft at a variety of ranges. The application scenario is a very close fly-by trajectory on a diverse set of asteroids. The considered targets are characterized by different shapes, sizes, and appearances, highlighting the generality of the proposed navigation pipeline. In-depth testing of the image-processing algorithms is carried out using high-fidelity synthetic images generated in a variety of geometric and illumination conditions. The performance of the navigation architecture is assessed through an extensive Monte Carlo campaign. Results show that the proposed pipeline provides accurate navigation in a wide range of conditions.