U. M. Schwuttke, A. G. Quan, R. Angelino, C. L. Childs, J. R. Veregge, R. Y. Yeung, and M. B. Rivera
Real-time AI is gaining increasing attention for applications in which conventional software methods are unable to meet technology needs. One such application area is the monitoring and analysis of complex systems. marvel (multimission automation for real-time verification of spacecraft engineering link), a distributed monitoring and analysis tool with multiple expert systems, was developed and successfully applied to the automation of interplanetary spacecraft operations at the Jet Propulsion Laboratory (JPL) of the National Aeronautics and Space Administration (NASA). In this chapter, we describe marvel implementation and validation approaches, the marvel architecture, and the specific benefits that were realized by using marvel in operations. Marvel is an automated system for telemetry monitoring and analysis. It has been actively used for mission operations since 1989. It was first deployed for the Voyager spacecraft’s encounter with Neptune and has remained under incremental development since this time, with new deliveries occurring every 6 to 10 months. marvel combines standard automation techniques with embedded rule-based expert systems to simultaneously provide real-time monitoring of data from multiple spacecraft subsystems, real-time analysis of anomaly conditions, and both real-time and non-real-time productivity enhancement functions. The primary goal of marvel is to combine conventional automation and knowledge-based techniques to provide improved accuracy and efficiency by reducing the need for constant availability of human expertise. A second goal is to demonstrate the benefit that can be realized from incorporating AI techniques into complex real-time applications.