A Fault Detection and Diagnosis Framework for Ambient Intelligent Systems

Ambient intelligence (AmI) systems are smart interactive systems that perceive their surroundings using sensors and act upon them using actuators. One of the most common applications of such systems is Smart Homes. In this context, the ambient system can offer a great level of dependability if it is able to exploit available sensor data in order to autonomously perform diagnosis. However, ambient environments are dynamic in a sense that components, in general, and actuators and sensors, in particular, can be added or removed from the system at run-time. This dynamicity raises new challenges not addressed in the state of the art of fault detection and diagnosis techniques. Unlike classical control theory methods, control-loops between ambient system components cannot be pre-determined at design time. In this paper we propose a new approach based on the modeling of physical phenomena, allowing one to use available resources to predict the values that are supposed to be read by sensors. Comparing the predictions and the real readings allows us to detect potential faults. Fault detection may be followed by fault isolation, which tries to identify the faulty component precisely.