Services pervasifs contextualisés : modélisation et mise en oeuvre
Dejene Ejigu · HAL (Le Centre pour la Communication Scientifique Directe) · 2007
Pervasive or ubiquitous computing aims to integrate computing and computing appliances into the environment rather than having computers as distinct objects. This can be realized through applications that adapt their behavior to every changing environment. Such systems need to ensure that the adaptive behavior experienced is useful, relevant, non-distracting and consistent with individual and organizational goals. Such adaptation needs proper capturing, management and reasoning of constantly changing context. Context capturing involves extracting relevant context data about selected entities in the environment. Context management deals with representation, aggregation, interpretation, storage and processing of context data. Context reasoning is the process of drawing inferences or conclusions (unknowns) from known facts using information from the various sources of context. The computationally intensive characteristics of context reasoning process, the presence of handheld or wearable, tiny and resource hungry computing devices, and the lack of a semantically rich context model have been a bottleneck for the development of such applications. Moreover, most of the current context-aware systems are based on ad-hoc models of context, which causes lack of the desired formality and expressiveness. They do not separate processing of context semantics from processing and representation of context data and structure. In this thesis, we propose a semantically rich and a collaborative context representation and management model that uses a hybrid of ontology and database management approaches (called HCoM model: Hybrid Context Management model). HCoM model uses ontology for modeling and management of context semantics and relational database schema for modeling and management of context data. These two modeling elements are linked to each other through the semantic relations built in the ontology. Separation of the two context modeling elements allows us to extract, load, share and use only relevant context data into the reasoner in order to limit the amount of context data in the reasoning space. By doing this, we considerably improve the performance of the reasoning process. The building blocks of the HCoM model are context data, context ontology, and deduction rules. These data elements are organized into a context representation structure (called EHRAM: Entity, Hierarchy, Relation, Axiom and metadata). EHRAM is a graphical context representation structure that serves as a context conceptualization model. EHRAM is mapped to a standard relational database schema for representation of its context component and is serialized to markup languages for representation of its ontology and rule component. We also present a domain independent context-aware middleware platform (called CoCA: Collaborative Context-Aware service platform) under which our proposed context management model is implemented and used. CoCA uses data organized into the HCoM model as its data source and provides reasoning and decision services based on changing contexts. It triggers proactive and/or reactive actions and provides a collaboration interface between the pervasive peers. CoCA collaboration is based on JXTA protocols and its reasoning is based on Jena framework. To evaluate the scalability and extensibility of the proposed model, reusability of the platform and performance of the collaboration process, we have developed a test case of the use of our context model in the platform using data from multiple scenarios: Community based network in a campus, smart hospital and adaptation of HCI to context. Results obtained from our experiment show that compared to other related works in the domain, our approach gives a robust, extensible and scalable model and platform for the development of context-aware applications in pervasive environment.