cr-models: An Inference Engine for CR-Prolog

CR-Prolog is an extension of the knowledge representation language A-Prolog. The extension is built around the introduction of consistency-restoring rules (cr-rules for short), and allows an elegant formalization of events or exceptions that are unlikely, unusual, or undesired. The flexibility of the language has been extensively demonstrated in the literature, with examples that include planning and diagnostic reasoning. In this paper we present the design of an inference engine for CR-Prolog that is efficient enough to allow the practical use of the language for medium-size applications. The capabilities of the inference engine have been successfully demonstrated with experiments on an application independently developed for use by NASA.