Abstract: A global path planning method based on genetic algorithm is studied with emphasis on how to find a path efficiently in large scale real terrain data set. This work is different from the previous work in two aspects. The first is the data set and the planned data is constructed using the real terrain data, including digital elevation data （DEM） and culture vector data. The vector data has topological information of points, lines and faces, including the culture such as area, river, bridge etc . The DEM and area data is organized using regular grid, the nodes of the grid store the elevation and area property. The river and road is organized used linked list. The advantage of the organizing method is its concision and convenience, and saving storage space. Second, domain knowledge is added into the framework of genetic algorithm, in the coding of chromosome, generating initial population and the operator of genetic algorithm. By adding these knowledge, the path planning problem can be solved more efficiently and effectively. The experimentation and the result are also given, and the result shows that the method improves the ability and efficiency of genetic algorithm in global path planning.