There is today plethora of communication-enabled computing devices that can be embedded in our environment for sensing and acting purposes. This opens the way to the implementation of advanced pervasive applications, providing assistance to human beings in their living places. In this research work, we focus on service-oriented platforms that are today heavily used to support such applications. These platforms have the ability to dynamically provide services, according to the applications needs. These services can present APIS of devices available in the environment like a thermometer or a door sensor, or more abstract functions providing the temperature in a room or higher-level security facilities. Pervasive applications share such services in order to achieve different and sometimes conflicting goals. These conflicts need to be managed in order to keep houses in consistent states.
This thesis defines an approach for managing conflicts between home applications in a service-oriented platform. This approach is optimistic and addresses conflicts at runtime via a causal model of the environment, called context. It is based on three points: the first one focuses on the description of conflicts in a context modeled as service-oriented components; the second point consist in the extension of the programming model of pervasive applications with locking/unlocking mechanisms and callbacks; the third point focuses on conflict management using a three-phase approach (prevention, detection and resolution). The proposed solution has been validated with the iCasa smart home platform. It has led to extensions of the iPOJO component model, which is the programming model of the pervasive applications but also of the iCasa context. This work has been carried out in collaboration with Orange Labs Meylan.