Robust, Adaptive Machine Learning (RAM)

With the increase in new foreign launches and the number of agile or potentially separable satellites being placed into orbit, it is critical that the U.S. Air Force not only maintain custody of these satellites, but also quickly identify and interpret unusual or unexpected behavior in complex and evolving situations. To support decision making processes and to facilitate rapid situational awareness, Numerica has initiated the development of an Adaptive Pattern-of-Life Learning Operations system (APOLLO) that exploits multiple intelligence (multi-INT) data streams to provide automated indications and warnings (I&Ws), event predictions, and associated confidence measures. APOLLO builds upon Numerica’s multi-INT change detection system, Athena, as well as Numerica's space surveillance tracking system, MFAST, and is being matured using historical and real-time data sources including data being produced by the Numerica Telescope Network (NTN). Through an innovative feature space representation of a satellite’s multi-INT spatial-temporal pattern of life, together with the application of powerful machine learning techniques, APOLLO supports a number of use cases of interest to the space battle management command and control (BMC2) community. In Phase II, APOLLO will be further matured using real data to improve robustness against variability in data quality, quantity, and cadence.