Description
Given the continual increase in the number of satellites in orbit, there is an escalating need to enhance trajectory models and operational risk management to gain a deeper understanding of orbital anomalies, such as unexpected changes in a satellite’s orbit, which can have a significant impact on satellite operations and threaten human and financial resources. This study aims to employ machine learning techniques to identify abnormal or divergent changes in satellite trajectories, from datasets containing multiple sets of Two-Line Elements (TLEs). The automatic detection of these behaviors has the advantage of avoiding the computation of satellite position and velocity. In this case, which is an unsupervised framework, we have selected the LOOP(Local Outlier Probability), One-Class SVM(Support Vector Machine), and Isolation Forest methods. We choose these methods for their different perspectives on density and distance. Once anomalies are detected, it is interesting to categorize them based on their causes using supervised machine learning methods, namely Random Forest, SVM, and artificial neural networks.
| Stream | Science |
|---|