SLAM (simultaneous localization and mapping) is widely used for selflocalization of robot, which is a basis for autonomous navigation of it. In order to autonomously navigate in unknown environments, a robot has to move around and generate the environmental map while localizing itself. However, SLAM in large scale environments is still a challenging research topic, and there is no complete system been reported yet that integrates SLAM and geographic information in such environments. In this paper, we propose an autonomous navigation system consisting of a GIS (Geographic Information System) database, SLAM system, and path planning method. The system is based on the computing capability of the GIS database. It can provide navigation guidance for a robot in a large geographic space. It is highly reusable and extensible. Moreover, the mapping result can serve as the basis of relocalization of a robot in the area even without a GPS sensor. Combining the geographic information and the mapping results of a robot is of great help for the robot to understand the environment, and in the meantime, to enrich the origin geographic information with more details.