Organizing large events raises security issues that require a better understanding of dense crowd dynamics. Until now, the focus in the field of pedestrian modeling has been on low or moderate densities. However, there is now an attempt to shift towards higher densities and larger crowds. This becomes even more challenging when one wants to include pedestrians' ability to anticipate how the rest of the crowd will behave, which necessitates addressing the problem within a game-theoretical framework. The aim of this internship is to identify and develop, for a couple of configurations of practical interest, game-theoretical models which are best adapted to the size and necessary degree of anticipation, so that the model contains just the right amount of anticipation in order to capture the main characteristics of crowd behavior without being unnecessarily costly in terms of computation. This work will be performed in collaboration with the Laboratoire Central de la Préfecture de Police (LCPP) which will provide relevant situations of interest to explore. The internship will have both a significant analytical component, involving understanding the properties of the different models and how they relate one to each other, and a strong numerical component.