Response time satisfaction in a real-time knowledge-based system

Abstract

Response to interrupts within a certain time frame is an important issue for all software operating in real-time environment. A knowledge-based system (KBS) is no exception. Prior work on real-time knowledge-based systems either concentrated on improving the performance of the KBS in order to meet these constraints or focused on producing a better solution as more time was allowed. However, a problem with much of the latter research was that it required inference-time costs to be hardcoded into the different branches of reasoning. This limited the type of reasoning possible and the size of the KBS. Furthermore, performing the analysis required to derive those numbers is very difficult in knowledge based systems. This research explored a model for overcoming these drawbacks. It is based on integrating conventional programming techniques used to control task processing with knowledge-based techniques used to actually produce task results. The C-Language Integrated Production System (CLIPS) was used for the inference engine in the KBS; using CLIPS for the inference engine simplified the rapid context switching required. Thus, the KBS could respond in a timely manner while maintaining the fullest spectrum of KBS functionality.