Abstract : Since 2001, in the Oasis team, I have developed research on the semantics of applications based on distributed objects, applying in the context of a real language, and applications of realistic size, my previous researches in the field of process algebras. The various aspects of this work naturally include behavioral semantics and the definition of procedures for model generation, taking into account the different concepts of distributed applications, but also upstream, static code analysis and model abstraction techniques, and finally model checking tools. In this thesis I show the complexity of this research and the wide variety of techniques involved. We have established a consistent semantic model based on a very flexible model called pNets (parameterized Networks of automata), which offers a good compromise between decidability, complexity, and usability. This approach allowed us to define semantics for behavioral aspects of object-based or component-based distributed applications, but also a notion of finite abstraction of the behavioral models, opening the use of model-checking tools. All these elements led to the realization of prototypes, in the Vercors platform, and case studies of large size.