Abstract: With the rapid increase of vehicles, the road network has become increasingly congested. It is of great significance to study the maximization of road network traffic efficiency, personal traffic time and path optimization. The existing methods periodically monitor congestion and replan paths. The previous path obtained by this kind of algorithm often leads to the detour of the next replanning, which is not conducive to the global optimization. This paper proposes a path planning algorithm based on congestion prediction. By analyzing the temporal and spatial correlation in a large number of historical traffic flow data, the traffic flow parameters at the future time are predicted, and then the congestion state classification is obtained through entropy weight method and fuzzy comprehensive judgment. A more reasonable path is planned by comprehensively considering the current and future congestion conditions. The simulation results show that compared with the dynamic path planning algorithm, this algorithm reduces the travel time by 5.35% and the travel distance by 11.72%.