Deep Learning for Navigation of Small Satellites About Asteroids: An Introduction to the DeepNav Project

CubeSats represent the new frontier of space exploration, as they provide cost savings in terms of production and launch opportunities by being able to be launched as opportunity payloads. In addition, interest in minor bodies is gradually increasing because of the richness and exploitability of the materials present throughout their surface, the scientific return they could yield, and their dangerousness. Moreover, they are characterized by a highly harsh environment. These are the reasons why greater autonomous capabilities are desirable for future space missions. Optical navigation is one of the most promising technique for retrieving spacecraft state, enabling navigation autonomy. Unfortunately, most of these methods cannot be implemented on-board because of their computational burden. This paper presents the “Deep Learning for Navigation of Small Satellites about Asteroids” project, in short “DeepNav”, whose aim is to change the current navigation paradigm by exploiting artificial intelligence algorithms for on-board optical navigation. As a result, DeepNav will evaluate the performance of fast and light artificial intelligence-based orbit determination for the proximity operations phase around asteroids.