Association cohérente de données dans les systèmes temps réel à base de composants - Application aux logiciels spatiaux

Distributed real time architecture of an embedded system is often described as a set of communicating components. Such a system is both data ow (for its description) and time-triggered (for its execution). This thesis ts in with these problematics and focuses on the control of the time compatibility of a set of interdependent data used by the components of the system. The architecture of a component-based system forms a graph of communicating components, where more than one path can link two components. These paths may have di erent timing characteristics, so information which transits on these paths takes various time to reach the nal component. However, the ows of information need to be adequately matched, so that the nal component uses inputs which all (directly or indirectly) depend on the same production step of the initial component. We call this property consistent data matching. The data matching property is de ned and formalized. A methodology is proposed to detect, in a component graph, the architecture con gurations that have to be analyzed. Several approaches are developed to manage data matching in periodic systems, without considering strict properties on the system scheduling. First, we consider systems composed by components sharing the same period and where intra-periodic communications are forbidden. Data matching is managed using queues that allows to balance the data transit times through the several paths. Then, we study systems where components have independent periods. Queues are also used and data timestamping is added to record data dependencies. Thus, a component is able, for each of its activation steps, to select consistent data sets according to data dependencies among the available input data. Data matching consistency is relaxed to allow the use of approximately consistent data sets. We use ltering queues which record only one data among a given number they receive. Their use allows to reduce the necessary queue size. Due to the loose execution model constraints, some situations exist where data matching liveliness is not guaranteed. Moreover, particular system architectures generate con ictual constraints and lead to an impossible data matching management. The thesis results are applied on the software of an earth observation satellite constellation, Fuego, which detects res or eruptions.