Abstract: Existing road network models fail to group conflicting lanes at junctions and calculate vehicles' locations efficiently in traffic animations.This paper presents a novel model-hierarchical semantic model of road networks to address these problems.It extends traditional descriptions of lanes and gives a broad road lane-Lane, which is subjected to certain constraints.We obtain Lanes according discretization and compression and then treat them as the atomic layer of the semantic model.We then build hierarchical semantic concepts to describe connection/adjacency/conflict relations among Lanes.Traffic phases on conflicting areas could then be generated efficiently to separate Lanes with different directions according to these semantic data.Experimental results show that:for the road networks within 5km from the downtown of a city, all of the semantic data and traffic phase information could be obtained automatically in 200 s.The adoption of such data can significantly improve the efficiency of location calculations in traffic animations and give a smooth trajectory presentation.