Abstract:
An expert system for satellite anomaly resolution must perform monitoring, situation assessment, diagnosis, goal determination and planning functions in real time. StarPlan is such a system being developed at the Ford Aerospace Sunnyvale Operation. This paper details the evolution of the StarPlan architecture from a rule-based system in which multiple "experts" classified and resolved anomalies to a more generic architecture that utilizes an object model of the domain to perform fault diagnosis using causal reasoning. The StarPlan I architecture is described; the lessons learned in StarPlan I implementation are discussed; and the architecture of StarPlan II is presented.