Artificial intelligence techniques applied to on-board space navigation, surveillance and tracking

Space debris constitute a huge threat for the actual and future space traffic. The constant monitoring and Resident Space Objects (RSOs) catalogue maintenance are essential to enhance the "space segment safety". Ground based networks of radar and optical sensors are not enough to face the evolution of such a risky phenomenon. By this, the idea of using already on-board star sensors for a fast, deployable and cost-effective constellation of space sentries. What if all this would be integrated with Artificial Intelligence (AI) techniques? This work presents an AI-based algorithm development for RSOs detection \& tracking within the Field Of View (FOV) of electro-optical attitude sensors. This can also be used for navigation functionalities such as High Angular Rate (HAR) determination in a quaternionless situation. The main images processing functions needed for this tasks will be discussed and faced through the AI and coupled with a developed tracking algorithm. Tests, comparison, tasks achievability with real and simulated images will be shown together with the used and trained Machine Learning (ML) models. In the end, foundations developments and keypoints ideas to develop a dual-purpose AI assisted autonomous Star Tracker (ST) for Space Surveillance and Tracking (SST)/ Attitude Navigation (AN) will be highlighted and presented.