©2001 IMAGESTATE A nomalies are unexpected conditions that occur in a functional engineering system. They must be detected, diagnosed, and resolved in order to maintain the system in its functional role. Managing anomalies in space systems is particularly chal- lenging given their complexi- ty and their remote orbital environment. In this article, we describe our work in developing a model-based theory of anomaly manage- ment, implementing it in the form of software algorithms, and applying it to the operation of a spacecraft and its distributed ground control network. Space systems are instrumental in generating, processing, and delivering a wide variety of products and services. A spacecraft’s global view, its location above the atmosphere, and its distinc- tive environment all provide unique characteristics that can be exploited for commercial, civil, and military applications. Satellites are routinely used to monitor the weather, to pro- vide communications services, to broadcast navigation signals, and to explore the solar system. Unfortunately, the very attributes that make space systems attractive also pose considerable challenges to their efficient operation. In particular, ensuring the health of a system and managing anomalous conditions is exacerbated by the extreme nature of the space environment and the need to remotely operate the system without the luxury of direct inspection. Exacerbating these issues is the complexity of modern spacecraft and their distributed ground support net- works, limitations on sensor information and configuration control, intermittent system connectivity due to orbital motion, and limitations on onboard resources such as power, communications bandwidth, and computational capability. Historically, system anomalies have been managed through the use of human-based “experiential” reasoning techniques. Highly trained and experienced engineers embed their com- partmentalized understanding, rules of thumb, intuitions, heuristics, and past experiences into a loose knowledge base composed of procedures, diagrams, handbooks, manuals, and remembered information. Widespread reports in the space operations literature, as well as years of the author’s own expe- rience in operating a number of space systems, attest to the significant drawbacks of this approach. Human-based experi- ential systems suffer from high training and staffing costs, sen- sitivity to personnel changes, the impacts of human error, the inability to reuse knowledge and procedures across lifecycle phases and missions, the sensitivity of the knowledge base to small changes in the system, and many other factors [1]–[3]. Together, these drawbacks can result in on-orbit operations costs that constitute 25–60% of overall mission lifecycle costs [4]; for the US$100 billion space industry, such system opera- tions costs range in the tens of billions of dollars annually [5]. Declining federal outlays for space projects and increased mar- ket pressures on commercial space ventures are forcing the space industry to lower these costs. As a result, new approach- es for detecting, diagnosing, and responding to system anom- alies are of great interest. Reasoning from First Principles To address the drawbacks of traditional experiential reasoning approaches, significant research has been performed over the past three decades in the field of model-based reasoning BY CHRISTOPHER KITTS Managing Space System Anomalies Using First Principles Reasoning Detecting, Diagnosing, and Resolving Problems That Occur in Satellites and Their Ground Networks © ARTVILLE DECEMBER 2006 IEEE Robotics & Automation Magazine 39 1070-9932/06/$20.00©2006 IEEE