The research of evacuation in some emergencies, e.g. fire, is of great benefit to reducing the injuries of persons. In this paper, a cellular automaton evacuation model based on mobile robots behaviors is presented. Each person is treated as an intelligent mobile robot, and motor schemas, including move-to-goal, avoid-obstacle, swirl-obstacle and nervous-motion, drive persons to interact with their environment. The motor schemas are combined with cellular automaton theory, and an evacuation model is built. Evacuation simulation of persons with different move velocities shows that the presented model can predict accurately the evacuation phenomena in some emergencies.