This presentation examines the use of knowledge-based and statistical agents for anomaly detection and root cause analysis in space systems. Statistical methods reliably identify deviations in high-rate spacecraft and ground system telemetry but provide limited diagnostic context. Knowledge-based methods, drawing on system models, interface documents, fault trees, and historical anomaly records, provide that context but scale poorly on their own.
The talk presents an architecture in which the two agent types operate together: statistical agents detect and characterize deviations, and knowledge-based agents reason over system relationships and operational history to produce a ranked set of candidate causes with supporting evidence. Coverage includes the interfaces between agent types, known failure modes of each approach, verification and human-in-the-loop requirements, and current limitations.
The intended audience is a mixed technical and business group. Attendees will leave with an understanding of the architecture, its applicability to operations workload and mission assurance