Distributed Consistency-Based Diagnosis without Behavior

Abstract

This paper faces the task of distributed diagnosis without exploiting any component behavioral model. A diagnosis problem is specified by an observation that just states whether each system output is either correct or incorrect. The system is split into parts, and a distinct diagnoser, which is supplied with knowledge that has been compiled off-line and is capable of communicating with its neighbors, is assigned to each of them. A family of methods is proposed to compute local and global minimal diagnoses that are consistent with both the observation and the system description, the latter being a kind of control flow structure.