This thesis is part of the ANR COMPER project.
The purpose of this thesis is to define a personalization meta-model, which will be used to implement personalized learning in three training areas related to the project. This meta-model should make it possible to propose a sequence with personalized activities that will be either adequate resources selected from existing activities or activities generated to meet the learner's needs. This personalization of activities according to the learner's objectives, behaviours and skills will be based on a skills framework and pedagogical strategies defined by the pedagogical team. In addition, the system will have to be able to observe its own operation based on the analysis the process traces to evaluate it and thus be able to improve the personalization process.