In this paper, we propose a position estimation method fusing geographic information and multiple sensor information, and a navigation method so that a mobile robot moves from a start position to a goal autonomously. We define two types of sensing points for correcting the self-position of the robot. One is the sensing point (GPS surveying point) based on GPS (Global Positioning System). The other is the sensing point (landmark sensing point) based on the stereo vision. We assume that the optimal path and the sensing points are given. At the landmark sensing point, the robot takes geographic information and estimates self-position with higher accuracy than GPS surveying value by using a stereo vision and a compass. In navigation phase, the robot uses geographic information and moves along planned path by dead reckoning with avoiding obstacles. In order to demonstrate the validity of the proposed method, experiments have been performed.