Abstract: In this paper, a functional vision approach with multiple scale sensors in space and time is proposed for the autonomous navigation of an intelligent mobile robot. The basic idea is to design special sensing and processing methods in different scales in order to carry out more effectively the required visual behaviors, such as road following, obstacle avoidance, global localization, etc.. It is emphasized that the functional vision is not necessarily based on high level understanding or 3D recovery of the environment, and the tasks of visual navigation can be fulfilled better by recognizing the functionality of the environment and objects in the environment. A comprehensive functional vision system has been designed by integrating fine scale binocular gaze images, large scale omni directional view images, and long time spatio temporal images. A unified multiple scale visual sensing system is proposed to combine binocular vision, omni directional vision and spatio temporal imaging of both sides into a single video camera.