Local Consistency and Junction Tree for Diagnosis of Discrete-Event Systems
Kan John Priscilla, Alban Grastien · Frontiers in artificial intelligence and applications · 2008
We extend the decentralised/distributed approach of diagnosis of discrete-event systems modeled using automata. The goal is to avoid computing a global diagnosis, which is expensive, and to perform local diagnoses instead. To still ensure global consistency, we transform the topology of the system into a junction tree where each vertex represents a subsystem. Local consistency between the diagnoses of these subsystems ensures global consistency due to the tree structure. This technique will work best for systems whose natural structure is close to a tree structure, as the generated automata will be of reasonable size.