Highly integrated domain-specific environments are essential for the efficient design of complex physical products. However, developing such design environments is today a resource-consuming error-prone process that is largely manual. Meta-modeling and meta-programming are the key to the efficient development of such environments.

The ultimate goal of our research is the development of a meta-modeling approach and its associated meta-prograrnming methods for the synthesis of model-driven product design environments that support modeling and simulation. Such environments include model-editors, compilers. debuggers and simulators. This thesis presents several contributions towards this vision, in the context of the Modelica framework.

Thus, we have first designed a meta-model for the object-oriented declarative modeling language Modelica, which faci litates the development of tools for analysis. checking, querying, documentation, transformation and management of Modelica models. We have used XML Schema for the representation of the meta-model. namely, ModelicaXML. Next, we have focused on the automatic composition. refactoring and transformation of Modelica models. We have extended the invasive composition environment COMPOST to handle Modelica models described using ModelicaXML.

The Modelica language semantics has already been specified in the Relational Meta-Language (RML). which is an executable meta-programming system based on the Natural Semantics formalism. Using such a meta-programming approach to manipulate ModelicaXML, it is possible to automatically synthesize a Modelica compiler. However. such a task is difficult without the support for debugging. To address this issue we have developed a debugging framework for RML, based on abstract syntax tree instrumentation in the RML compiler and support of efficient tools for complex data structures and proof-trees visualization.

Our contributions have been implemented within OpenModelica, an open-source Modelica framework. The evaluations performed using several case studies show the efficiency of our metamodeling tools and methods.