Abstract: Because of uncertainties of the environment and the robot mathematic model and some limits of these robots,it is difficult for autonomous micro mobile robots to avoid obstacle. In this paper,we design an automatic avoid-obstacle avodance behavior. This behavior uses the artificial motor neuron network to select its moving mode and uses reinforcement learning to adjust the weight of the network at real time. The results of the safe wander are made of wander behavior,and automatic obstacle avoidance behavior have shown its validity.