Abstract : This work is part of the AROS project. Its goal is to define a fast prototyping tool for dynamic and distributed real-time applications, mostly for automotive industry and robotic. Two distinct methods are normally used to develop distributed real-time applications. The first one -the time triggered approach- is based on worst execution time analysis, whereby time sharing for the various tasks of an application is statically defined. This approach offers considerable safety but the time analysis is sometimes difficult to process. The second one -the priority scheduling approach- is based on ascribing a priority level to each task, which will then allow the system to define an execution order, based on the events is has received. This second approach is more flexible and easier to implement but is less safe and cannot ensure that the application behaves predictably. The structure of the AROS applications being dynamic, the time-triggered approach is irrelevant as it requires a static analysis that cannot be conducted. The priority scheduling approach is also irrelevant because of the non predictable behaviour. We propose an approach based on causal events scheduling inspired by distributed event simulators scheduling techniques. While comparatively easy to use for application designers, this new approach produces applications with a perfectly predictable behaviour. Two main obstacles must be overcome: the real time synchronisation of the execution engine and compliance with real-time constraints.