Abstract: A new fuzzy dynamic traffic assignment (FDTA) model was proposed. Firstly, the fuzzy theory was used to describe the real-time link travel time and path travel time. Secondly, fuzzy shortest path sets were found based on h-level sets, and then their membership functions were calculated. Thirdly, the C-LOGIT model was used for network loading. At last, FDTA model was compared with classical stochastic dynamic traffic assignment (SDTA) model using a small traffic simulation network. The simulation results show that FDTA model can achieve more reasonable network traffic flow, and theoretically resolve the problem of traffic network dynamic traffic flow modeling as the congestion, traffic incidents or construction happened. Further, when the traffic network becomes complex, by changing the value of h, the number of fuzzy shortest paths could be decreased and the computation efficiency could be improved greatly.