A Framework for Decentralized Qualitative Model-Based Diagnosis

In this paper we propose a framework for decentralized model-based diagnosis of complex systems modeled with qualitative constraints and whose models are distributed among their subsystems. We assume that Local Diagnosers are associated with subsystems and are coordinated by a Supervisor which acts as the diagnoser for the complex system. The Local Diagnosers and the Supervisor communicate via a standard interface and share a common modeling ontology. In this diagnostic architecture, connections between subsystems only need to be known at runtime, thus allowing for dynamic (re)configuration of the system. The approach is designed to compute partial hypotheses in order to avoid unnecessary queries to Local Diagnosers.