Fine-grain adaptation for real time embedded systems using UML/MARTE profile

The integration of adaptation techniques in the development cycle of real time embedded systems (RTES) has become a necessity to permit their fast and efficient reaction to environment variability. In order to yield considerable benefits in terms of resource utilization, the adaptive behavior should be integrated in different system layers and with different granularities (fine and coarse). The existing model-based design approaches of adaptive RTES is limited to the modeling of coarse-grain adaptation techniques which bring global modification to the whole system configuration. However, fine-grain adaptation techniques, specifically per-application adaptation, which is an important ability of embedded systems for dealing with modern applications complexity, are still under-explored. Since this adaptation issue is not supported by the UML MARTE profile, we propose in this paper a model-based design of an adaptive application layer using the MARTE standard. We define an additionnal package extending the Software Resource Modeling sub-profile. A case study on the H264/AVC video encoder application is presented to illustrate the application of our proposed extension.