Applying Machine Learning to Routine Satellite Ground Segment Operations by Means of Automated Anomaly Detection

Abstract

To tackle the domain specific challenges spacecraft operations poses on anomaly detection methods, the Automated Telemetry Health Monitoring System (ATHMoS) was developed at the German Space Operations Center (GSOC) and integrated into our Visualisation and Data Analysis software (ViDA). The main challenges include the peculiarities of the telemetry data transmitted by the satellites, the limitation of resources and accuracy and usability requirements posed by the end users. The ATHMoS was designed with these challenges in mind and uses sets of generic statistical properties to derive an explainable anomaly probability.