Multistage learning applied to obstacles avoidance is studied in this paper. We propose a new learning system which consists of the hierarchical fuzzy rules, fuzzy evaluation system and 2-stage learning automata. Then we show how an autonomous mobile robot can acquire the optimal action and fine attentive behavior using multistage learning through the interaction with the real world. In other words, the robot acquires how to pay attention to moving obstacles and how to avoid them using the steering and velocity control inputs, simultaneously. We also show the experimental results to confirm the feasibility of our method.