Applying knowledge compilation techniques to model-based reasoning

摘要

The application of knowledge compilation techniques to a specialized class of knowledge-based systems known as model-based reasoning (MBR) systems is discussed. Knowledge compilation is defined and explained, and the motivation for applying knowledge compilation techniques to MBR systems, namely, to improve their run-time efficiency on tasks such as diagnosis, design, and simulation, is examined. As an example, a prototype knowledge compilation system capable of compiling both diagnosis and design rules from a general-purpose structure/behavior model of the reaction wheel assembly of NASA's Hubble Space Telescope is considered. R. Davis' criticisms (see proc. 1989 workshop on model-based reasoning, 11-27) and the author's response are reviewed.