Systematic Methods for Knowledge Acquisition and Expert System Development

Nine cooperating rule-based systems, collectively called AUTOCREW, were designed to automate functions and decisions associated with a combat aircraft’s subsystems. The organization of tasks within each system is described; performance metrics were devel oped to evaluate the workload of each rule base and to assess the cooperation between the rule bases. Each AUTOCREW system is composed of several expert systems that perform specific tasks. AUTOCREW’s NAVIGATOR was analyzed in detail to understand the difficulties involved in designing the system and to identify tools and methodologies that ease development. The NAVIGATOR determines optimal navigation strategies from a set of available sensors. A Navigation Sensor Management (NSM) expert system was systematically designed from Kalman Filter covariance data; four ground-based, a satellite based, and two on-board INS-aiding sensors were modelled and simulated to aid an INS. The NSM Expert was developed using the Analysis of Variance (ANOVA) and the ID3 algorithm. Navigation strategy selection is based on an RSS position error decision metric, which is comp uted from the covariance data. Results show that the NSM Expert predicts position error correctly between 45% and 100% of the time for a specified navaid configuration and aircraft trajectory. The NSM Expert adapts to new situations, and provides reasonable estimates of hybrid performance. The results also show how the NSM Expert chooses the best or next-best navigation strategies in situations where computational resources are limited. The systematic nature of the ANOVA/ID3 method makes it broadly applicable to expert system design when experimental or simulation data is available.