Marta Guimaraes, Neuraspace; Claudia Soares, FCT-UNL; Chiara Manfletti, Neuraspace
Keywords: Space Situational Awareness, Space Debris, Unsupervised Learning, Clustering
Abstract:
In recent years, the number of objects orbiting in the low Earth orbit (LEO) has witnessed a significant increase, which poses a challenge to decision-making and overall space situational awareness. To make critical decisions such as collision avoidance manoeuvres, satellite owners and operators must be aware of the objects in close proximity to their assets. To achieve this, it is crucial to understand the characteristics of such resident space objects (RSOs). In this study, we analyse a large dataset of catalogued objects and publicly available two-line element data to identify the main characteristics of RSOs in LEO. We use unsupervised learning techniques to find patterns in the data and group the RSOs by their intrinsic characteristics, orbital information, and patterns of life. Specifically, we use different clustering algorithms to identify essential clusters leveraging both the static and the dynamic information of the RSOs. We further explore the obtained clusters by using dimensionality reduction techniques, providing deeper insights into the dynamics of objects in orbit. We discuss the limitations of current satellite datasets in providing comprehensive information, such as the level of automation of RSOs or classification criteria for new satellite shapes. Our work contributes to a better understanding of RSOs in LEO by uncovering patterns and identifying fundamental clusters of RSOs based on their characteristics, therefore improving space situational awareness and decision-making for satellite owners and operators.
Date of Conference: September 17-20, 2024
Track: Machine Learning for SDA Applications