Threshold-based context analysis approach for ubiquitous systems

International audienceIn ubiquitous computing systems, applications must be able to respond to dynamic context changes in order to provide suitable services to users. A promising solution consists of developing context aware applications which automatically change their behavior according to the user needs, the user preferences, the available resources and the surrounding environment. Furthermore, a context aware application is characterized by a closed feed back loop (MAPE) with four phases: Monitoring, Analysis, Planning and Execution. In this paper, we focus on the second phase of the MAPE loop. We elaborate a layered approach composed of a context capture layer, a context management layer and a context adaptation layer. Our approach enables systems to become pervasive in a transparent manner handling context monitoring and context analysis. We focus our work on the second layer (i.e context management). We propose first a context classification which takes into acount the context parameters evolution behavior. Then, we propose a context analysis approach for ubiquitous systems which aims at analyzing context information and detecting significant changes. The proposed approach uses a threshold based technique in order to detect context changes. When relevant context changes are detected, the context aware application will be notified to trigger its suitable process dynamically in order to deal with the changes