Cost-based abduction using binary decision diagrams

Abstract

This paper proposes an abductive reasoning system, which can find most preferable solution efficiently, using Binary Decision Diagrams. We propose a specialized BDD and its operation suitable for abductive reasoning: PBDD (Partial BDD) and GPC (Graft & Pruning Construction). We have implemented PBDD and GPC algorithm and built a cost-based abductive reasoning system which can find much more efficiently the most preferable explanation of a given observation. We have also made some experiments on the system with some diagnostic problems. Some good performance results are also shown.