Maintaining continuous dependability in sensor-based context-aware pervasive computing systems

Sensors are often indicative of events of interest, or events which a system should be aware of and should react to. In particular, such indications may reflect unsteadiness or instability with respect to the current contextual situation. This paper discusses the concepts of stability and instability in context-aware systems, and suggests a model, methodology and an algorithm for real-time analysis of sensor readings in order to estimate future changes in context as well as in assisting to maintain the proper functioning of context-aware pervasive systems. We provide a conceptual analysis of our approach, implementation details and experimental evaluation of simulated scenarios that are part of a motivating use case.