The main research topics of the DREAM project-team are about aiding monitoring and diagnosis of time evolving systems. The main issue is to infer the state of a system from observations provided by sensors in order to detect and characterize potential anomalies or failures within the system. We use a model-based approach relying on normal and faulty behavioral models. These models are temporal qualitative discrete-event models such as temporal communicating automata, temporal causal graphs or sets of chronicles.
- Diagnoser algorithm design and implementation.Relying on model inversion and compilation, these techniques aim at computing compact models which link directly observations to faults. More precisely, we focus on a decentralized and generic approach and on the use of model checking techniques.
- Diagnosis and decision interaction in an uncertain context.
- Automatic model acquisition. We investigate symbolic machine learning techniques such as ILP (Inductive Logic Programming) and temporal data mining.
- Environmental protection: qualitative modelling of pollutant (pesticides, nitrates) transfer in groundwater.
- Industrial diagnosis: supervision of telecommunication networks and power distribution systems.
- Health monitoring: electrocardiogram on-line analysis, health monitoring of big dairy herds.
International and national relations
- Collaboration with Rennes Hospital (Pharmaco-epidemiology Departement)
- Collaboration with ANU (the Australian National University, Canberra). Contact: Sylvie Thiébaux.
- Collaboration with INRA (Institut National de la Recherche Agronomique – National Institute for Agricultural Research) in several projects.
- Collaboration with Agrocampus Ouest – Rennes.