Real-time Diagnosis of Ambient Environments Using a Modeling of Physical Effects Combined with Temporal Logic
Résumé
Ambient intelligence systems interact with their surroundings using actuators and based on environmental data collected from sensors' readings. Diagnosis in this context must address some particular challenges due to the dynamic nature of these systems and the impossibility to pre-define control loops between sensors and actuators at design time. A possible solution to this problem is to base diagnosis on observed physical phenomena (effects) induced by actuators and to reason over a pre-defined ontology allowing one to apply physical laws, to compare calculated values with actual sensors' readings and thus to notice anomalies which corresponds to probable faults. This "effect"-based model, which describes the expected physical effects of the actuators onto the environment, allows one to perform basic diagnosis, using a static view of the system. However, to perform more complete diagnosis, we claim that one has to take the dynamics of the system into account. To achieve this, this paper proposes to extend the simple "effect"-based model with a behavioral model using temporal logic.