Laying a Foundation for Software Engineering of Knowledge Bases in Spacecraft Ground Systems

In spacecraft telemetry expert systems technology i s being used to manage the complexity generated by the increasing number of complex measurands. However, an uncontrolled proliferation of rules in an expert system can lead to maintenance and management problems of the system. A semi-automated tool, such as Pragati's MVP-CA (Multi-ViewPoint Clustering Analysis) tool, can provide a valuable aid for comprehension, maintenance, verification, validation, integration and evolution of these expert systems by structuring a large knowledge base in various meaningful ways. The MVPCA tool “mines” the knowledge existent in telemetry rule bases by exploiting the similarity across rules. This knowledge can serve as a handle to verify and validate the knowledge in the existing system as well as to formulate new rule sets for future mission planning activities. Introduction The increased number and complexity of spacecraft mission measurands and the evolution of ground systems architectures that support multiple configurable roles have emphasized the need to alleviate the mission operator workload. Rule-based expert systems are a common technology used to manage this complexity; yet a rule set created for a particular mission is often developed in a stand-alone, ad hoc manner. The consequence of this practice is that rule-based systems are redeveloped each time the system changes [Alvarado 1998]. Moreover, due to the critical nature of these applications, much more stringent standards have to be imposed now on their ability to provide reliable decisions in a timely and accurate manner. Pragati's Multi-ViewPoint-Clustering Analysis (MVP-CA) tool provides a framework for clustering large, homogeneous knowledge-based systems from multiple perspectives [Mehrotra & Wild 1995]. It is a semi-automated tool allowing the user to focus attention on different aspects of the problem, thus providing a valuable aid for comprehension, maintenance, verification and validation (V&V), integration and evolution of knowledge-based systems. The MVP-CA tool has recently been adapted for clustering telemetry knowledge bases. We present here some preliminary results of applying the MVP-CA tool on some telemetry expert systems. In particular, results exposing verification and validation (V&V) problems in the rule bases are presented. We will also briefly discuss our next step of extracting reusable components in a systematic manner by proposing an integration of the MVP-CA tool with case-based reasoning (CBR) technology. Issues relating to indexing, retrieval and adaptation of the rule sets can be addressed effectively when the two technologies are integrated. Motivation Expert systems are increasingly being used as intelligent information specialists in cyberspace, both for civilian and military applications. In spacecraft telemetry, expert systems technology is used to manage the complexity generated by the greater number of complex measurands [Lindsay 1998]. Spacecraft satellite telemetry (sub) systems have a unique characteristic in that they usually have multiple configurable roles; hence, there are similar rule bases in existence for different subsystems. As new missions get planned the number of such rule bases with similar structures keeps growing. Also, as new knowledge evolves due to new technology in the market, these systems have to be adapted to incorporate/reflect the changes in technology. Each mission has its own rule set to be applied and each one of them has the potential to grow into a monolithically large unmanageable system. The phenomenon of “add a rule each time” to take care of different situations in any expert system, leads very quickly to an uncontrolled proliferation of rules in the expert system. Due to the data-driven nature of expert systems, as the number of rules of an expert system increase, the number of possible interactions between the rules increases exponentially. The complexity of each pattern in a rule compounds the problem of management of rules even further. Documentation has the danger of becoming obsolete very quickly, as software developers do not always have the necessary discipline to keep updating their documentation. Furthermore, defining any requirements or specifications up front in such a rapid prototyping and iterative development environment, even though they are desirable, becomes an impractical and moot question. Even if they were specified, as any From: Proceedings of the Twelfth International FLAIRS Conference. Copyright © 1999, AAAI (www.aaai.org). All rights reserved. software, conventional or knowledge-based becomes more complex, common errors are bound to occur through misunderstandings of specifications and requirements [Bellman & Walter 1988]. It is therefore desirable to have an analysis tool that exposes a developer to the current software architecture and semantics of the knowledge base in such a dynamically changing development environment, so that the knowledge base can be comprehended at various levels of detail. To achieve this goal, the knowledge in the system has to be suitably abstracted, structured, and otherwise clustered in a manner that facilitates software engineering activities [Jacob & Froscher 1990, Landauer 1990]. Hence, by exposing the knowledge contained in the knowledge-based system through the Multi-ViewPoint Clustering Analysis tool, we formulate a basis for addressing reusability, maintainability, and reliability issues for such systems. Multi-ViewPoint Cluster Analysis (MVPCA) Existing approaches to structuring systems are limited in a major way. They only provide a single viewpoint of a system. We believe that no one single structuring viewpoint is sufficient to comprehend a complex system. In this paper we show the feasibility of applying Pragati’s Multi-ViewPoint-Clustering Analysis (MVPCA) methodology on satellite telemetry rule-based systems for V&V and reusability. MVP-CA framework also has the potential to be extended to incorporate casebased retrieval and adaptation technology for reusability of clusters generated through the MVP-CA tool. Analysis of Existing Rule Sets We have analyzed the following three telemetry rule bases: • Spacecraft Environmental Anomalies (SEA-ES) • X-Ray Timing Explorer (XTE) • Unexpected Events System Rule Base (UES). A brief discussion of the results follows. Spacecraft Environmental Anomalies (SEA-ES) Spacecraft Environmental Anomalies (SEA-ES) is an expert system developed by The Aerospace Corporation, Space and Environment Technology Center for use in the diagnosis of satellite anomalies caused by the space environment. The satellite anomalies to be detected by the rule base ranges from surface charging, bulk charging, single-event effects, total radiation dose, and space-plasma effects. Various parameters play a role in the determination of these anomalies such as, orbit of the satellite, the local plasma and radiation environment, satellite-exposure time, hardness of the circuits and their components etc. Pattern(s) : INCL became stable in group: Rule# Rule Description 138 INCL BT -5 5 => INCLIN = EQTRL 139 INCL BT 5 30 V INCL BT -5 -30 => INCLIN = L_INCLIN 140 INCL BT 30 60 V INCL BT -30 -60 => INCLIN = I_INCLIN 141 INCL BT 60 80 V INCL BT -60 -80 => INCLIN = H_INCLIN 142 INCL BT 80 90 V INCL BT -80 -90 => INCLIN = PLR 173 PERIGEE BT 96 145 ^ APOGEE BT 320 480 ^ INCL BT 80 100 => ORBIT = LOW1 175 PERIGEE BT 200 300 ^ APOGEE BT 200 300 ^ INCL BT 45 70 => ORBIT = STS57L 177 PERIGEE BT 280 420 ^ APOGEE BT 280 420 ^ INCL BT 45 70 => ORBIT = STS57H 179 PERIGEE BT 480 720 ^ APOGEE BT 480 720 ^ INCL BT 45 70 => ORBIT = ERBS 174 PERIGEE BT 135 205 ^ APOGEE BT 185 276 ^ INCL BT 85 105 => ORBIT = LOW2 180 PERIGEE BT 630 770 ^ APOGEE BT 630 770 ^ INCL BT 90 105 => ORBIT = L^SAT 181 PERIGEE BT 735 900 ^ APOGEE BT 750 920 ^ INCL BT 90 110 => ORBIT = DMSP 183 PERIGEE BT 800 980 ^ APOGEE BT 82