Section: Overall Objectives

On-line monitoring issues

Classical model-based diagnosis methodologies have been shown to be inadequate for complex systems due to the intractable size of the model and the computational complexity of the process. It is especially true when on-line diagnosis is considered and when the system is composed of a number of interacting components (or agents). This is why we focus on a decentralized approach which relies on computing local diagnoses from local models and synchronizing them to get a global view of the current state of the system. The problems we are investigating are: Which strategy to select for synchronizing in an optimal way the local diagnoses in order to preserve the efficiency and the completeness of the process? Which kind of communication protocols to use? How to improve the efficiency of the computation by using adequate symbolic representations such as BDD and partial order reduction techniques? How to ensure an efficient incremental process, in an on-line diagnosis context where observations are incrementally collected? How to deal with reconfigurable systems, the topology of which can change at runtime?

The diagnosis task is usually a steady task which does not take into account the current context and the possible repair actions. However, designing an adaptive system requires to define monitoring and diagnosis tasks which adapt to the context. A first possible improvement is to have a more active diagnosis, capable of tuning the observability of the system, deciding for instance to activate other sensors to acquire more information. A second idea is to adapt the diagnosis granularity, i.e. the details of the information that is needed, to the available repair procedures in the system.