Satellite Unknown Anomaly Resolution Using Cased-Based and Model-Based Reasoning
Stottler Henke developed a ground-based Unknown Anomaly Resolution System (UARS) for the U.S. Air Force at Schriever Air Force Base. Stottler Henke contracted with Raytheon to obtain access to anomaly resolution experts and models of satellite systems. The UARS employs both case-based reasoning (CBR) and model-based reasoning (MBR) in concert to diagnose and resolve unknown anomalies in engineering telemetry from geosynchronous communication satellites. An unknown anomaly is a set of anomalous telemetry mnemonics that do not match any previously encountered and documented anomalies. These might be instances of previously encountered anomalies that are manifest in an unfamiliar way or of truly novel anomalies. CBR prioritizes and prunes the search space that MBR must search in order to uncover potential ultimate causes for the anomaly. It also suggests possible resolutions once the fault has been identified. Recovery planning was based on an automatic, intelligent planning and scheduling assistant. Model-based reasoning was employed to enable the planner to determine: 1) how the satellite’s state was likely to evolve (with and without planned recovery actions; and 2) what risks the evolution entailed that the recovery plan would have to mitigate. The recovery planning process iteratively suggested and evaluated recovery plans until a satisfactory plan was achieved.