Active obstacle detecting and avoiding for autonomous robot are significant research fields.In the past

the vision system and motor controlling system of the robot have been developed independently.Most people ignored the relationship of these two systems

separated the close coupling

and reduced the ways to obtain more information.Here

a Vision-Action model is developed to describe the relationship between the robot vision and motor action

in which the coupling provides more information resources of the environment.In addition

reinforced learning is used for autonomous robot to avoid the dynamic obstacle based on this Vision-Action model.Our experiments in the method were carried out stably and in real time.