According to the characteristic that fluid can naturally explore reasonable flow direction and feasible path in complex environment, a global path planning method for autonomous driving based on virtual flow field is proposed. Firstly, a mathematical model of virtual road flow field is established based on the theory of fluid dynamics, and the computational fluid dynamics method is adopted to numerically solve the state of flow field, so as to generate multiple streamlines from the starting point to the destination in a large-scale complex road network. Meanwhile, by constructing a comprehensive evaluation function of the planned path, the optimal global path is selected from multiple streamlines. Finally, a combined GNSS/INS positioning system is designed to collect real data of road network and build a global static road map, and the proposed method is experimentally verified in a campus scenario. The experimental results show that the global path planning method based on virtual flow field can plan and select a complete and smooth global optimal path according to the automatic driving task and avoid the non-solution situation that falls into the local minimum of the environment.