With its imitating the perception and behaviors, scheming for instance, of human beings in virtual space, the crowd simulation technique aims at a truthful reflecting of crowd characteristics. However, for the crowd simulation, the behavior control is the crux. It introduces functions like perception, cognition, decision-making and so forth into motion control, and builds an autonomy model for virtual human, thus achieving the control of virtual crowd. To begin with, this paper introduces theories of behavior and behavior control. Based on it, the thesis presents a behavior control model, which involves perception model, path planning and behavior model. It goes on with the comparison of several behavior control models, and concludes the related control algorithm. Besides, based on the analysis of relevant researches, this paper ends with the indication about the aspects that more attention should be laid to in the field of crowd simulation in the near future.