Real-time systems can be large, distributed and have a dynamic environment. This requires the introduction of various operating modes and reliability techniques. Different operating modes are associated with a different architecture and behavior. Dynamic changes between these modes have an impact on the temporal characteristics of systems which requires an analysis whether the constraints of the system are also fulfilled during adaptations. The work presented in this thesis is focused on specifying the adaptability and the study of the temporal impact of dynamic adaptation actions on a predefined set of configurations. For this purpose, we present a tooled methodology based on the concept of Mode of the MARTE profile. Each mode represents a possible behavior of the system for a well determined environment associated with a software configuration. The influence of these operations on the temporal behavior of the system is done via schedulability analysis. This methodology proposes to model the adaptive behavior through the definition of the context, the variability, the reconfiguration operations and of the base configuration. The schedulability analysis is performed at the model level by incorporating the impact of the behavior of adaptation. Two paradigms of modeling can be exploited to perform this analysis: request/reply and data flow. This allows to verify that the temporal constraints of our system will remain satisfied even with the inclusion of reconfiguration operations executing the adaptive behavior. Finally, the approach allows generating the implementation of adaptive behavior from the model to automate the integration of adaptation mechanisms in real-time systems.
