Vehicle scheduling of urban distribution is an important element of the logistics center's operation. The total scheduling cost is not only influenced by the satisfaction of customer service time window and the effective utilization of vehicle capacity, but also is closely related to the traffic flow, road capacity, vehicle saturation degree of intersection, traffic lights and road rules in actual road network. This paper proposes a two-stage model based on the shortest path's selection in the road network and vehicle scheduling, and designs a two-stage algorithm to solve it which combines both Dijkstra algorithm with improving labeling rules and simulated annealing algorithm with parallel local search rules. By taking a distribution center in Shahekou District of Dalian as an example, a vehicle scheduling plan in two weeks was calculated and compared with the original plan, which verified the effectiveness of the model and algorithm. Sensitivity analysis of two road network factors: traffic flow and forbidden roads, showed that change of traffic flow on the main road and moderate distance between forbidden roads and distribution centers have a great effect on vehicle schedules.