FLAPPER: Fault Learning Agent for Predication, Protection and Early Response

Current fault detection and correction techniques require either an operator-in-the-loop to identify and respond to on-board satellite anomalies or a hard-coded, rules-based fault response tree to algorithmically respond to triggers and perform corrections or escalate the fault. Both processes consume time and resources from system engineers or ground operators and are unlikely to identify novel patterns in onboard data and telemetry that signify a fault event. Orbit Logic proposes the Fault Learning Agent for Prediction, Protection, and Early Response (FLAPPER) solution, to be implemented as a pair of Specialized Autonomy Planning Agents (SAPAs) that expand our onboard Autonomous Planning System (APS) architecture to include machine learning capable of detecting, isolating, and mitigating anomalies in real- or near-real-time with minimal ground intervention. FLAPPER will analyze a subset of onboard spacecraft health and safety data and telemetry to train against and later autonomously detect and categorize spacecraft faults. Categorized faults will then be mapped to acceptable corrective action responses to be carried out autonomously or with expert oversight from operators given the inferred data. Transitioning the fault detection and correction capabilities to an autonomous and onboard application will benefit the missionrsquo;s success by reducing the time the spacecraft spends in anomalous states.