Abstract: According to the requirements of unmanned ground vehicle(UGV) autonomous navigation, an implementation method was proposed for path optimization with given task points as constraints. Firstly, a geographic information system (GIS) database was built based on the GIS platform to provide priori information for UGV. And then a road-matching algorithm was designed based on the computational geometry. Simultaneously the static and dynamic path planning was conducted combining with A^* algorithm. Secondly, a new crossroads model was proposed according to the movement characteristics of UGV and the demand in recognizing the intersection. An algorithmic design for the intersection tracks and U-turn tracks was conducted to smooth the UGV's driving routes and to ensure the accuracy of intersection recognition. At last, a dynamic re-planning driving strategy was put forward. The validity of designed algorithm was verified by actual car experiments.