The simulation of human behavior raises several problems (navigation, perception, collision, animation). One of the problems present in the main quest for realism is to adjust the animation of each character according to its intent, its environment and its mood. While the techniques of motion capture has led to a strong improvement in quality, the techniques of animations adding are a pretreatment techniques depended on human intervention. The animations in real-time applications are not that much varied and reflect only a small portion of the virtual characters secondary intentions, its state of mind or mood. On the other side, the innovative techniques in AI as the one developed by SpirOps allow to model complex intentions, or even the feelings of virtual characters. By combining the know-how of the team SAARA in terms of character animation and SpirOps in term of AI, this thesis will seek to enrich the state of the art animation by integrating the behavioral simulation in the control of movements. The objective is to propose techniques to adapt realistically and on the fly the movements of each individual relative to its intent, its environment and its mood.