It is an important issue to make a disaster reduction plans for overcrowded population in the urban cities. In this paper, we focus on evacuation guidance to prevent the damage from spreading. The personal navigation system cannot deal with evacuation guidance for the human crowd with large numbers of individuals because of time constraints and extraordinary communication error. An implicit guidance based on dynamical characteristic of swarm behavior is efficient and effective with a few guidance operators. We propose a modeling and control method of swarm based on vector field. The evacuee behavior model contains intention of evacuation, field of view, collision avoidance and evacuee group, which are represented by vector field. The guidance operator model contains indicating direction. By giving the desired vector field that indicates the safe route for evacuation, the position of guidance operators are optimally distributed. Moreover, the number of guidance operators is minimized based on the contribution index. The proposed modeling and control method is applied to the swarm robot and the effectiveness is evaluated by the experiments.