Improving the scalability of rule base verification using binary decision diagrams: An empirical study

Abstract

As their field of application has evolved and matured, the importance of verifying knowledge-based systems is now widely recognized. Nevertheless, some problems have remained. In recent work, we have addressed the poor scalability to larger systems of the ATMS-inspired computation methods commonly applied to rule-chain anomaly checking. To tackle this problem, we introduced a novel anomaly checking method based on binary decision diagrams (BDDs), a technique emanating originally from the hardware design community. In this paper, we present further empirical evidence of its computational efficiency on real-life rule bases. In addition, we will investigate the issue of BDD variable ordering, and its impact on the efficiency of the computations. Thereby, we will also assess the utility of dynamic reordering. © 2004 Springer-Verlag.