We are used to interacting with other humans on a daily basis. These can vary from individual people to groups to large crowds and display behaviors on a wide range of scales. Crowd simulation, by modeling all these different interactions (though usually inter-pedestrian relations), aims to offer a better understanding of the mechanics behind these daily behaviors as well as reproduce the resulting, fascinating large-scale patterns. This report first gives a state-of-the-art of existing crowd simulation models and the little work that has been done to evaluate them. It then presents the mechanics behind the framework: ways to use the data, which models have been integrated, the metrics that have been designed and the algorithms that have been defined to calibrate the models. Finally, before concluding, the last chapter aims to describe the resulting framework as well as the preliminary results that have been obtained by using it.