The problem of automatically generating a world map for mobile robots is addressed. A proposed map, based on the cognitive map of human, is qualitative and contains information such as action at an intersection with a landmark, relative-location between two locations and a sequence of landmarks in a path. A method of pathfinding using the map and modes of robot movement necessary for tasks such as map generation and pathfinding are described. As a result of computer simulation, the validity of our approach have been certified.