The aim of this thesis is to develop techniques for the evaluation of the performance of algorithms for specific problems of scheduling. The work will consist in proposing experimental protocolsto assess specific problems of scheduling by restricting themselves to those who share similar instances as graphs. This implies on the one hand analyzing the problems considered in order to define what can berelevant instances. Next, it will be necessary to propose generators (random) of (difficult) instances, then perform an experimental analysis of existing algorithms.L'objectif du travail à réaliser dans cette thèse consiste à mettre au pointdes techniques pour l'évaluation de la performance d'algorithmes dans lecadre de l'ordonnancement. Le tra...