In recent years, drone scheduling has received increasing attention from researchers. In this study, a vehicle routing problem with drone stations is proposed, and an arc-based optimization model is developed with an objective of minimizing the total cost, in which various constraints such as the latest arrival time of packages, the limited number of drones and the maximum flying duration of drones are considered. By using the Dantzig-Wolfe decomposition, the arc-based model is decomposed into the path-based main problem model, the truck sub-problem model, and the drone sub-problem model. A branch and price algorithm is designed to solve the problem and to obtain the global optimal solution. Extensive numerical experimental results show that our proposed branch and price algorithm outperforms the mainstream commercial solver in terms of solution time and solution quality. In addition, the truck travel cost accounts for a high proportion of the total cost. Moreover, compared with the number of drones and the speed ratio of drones to trucks, the proportion of customers accessible by drones has a more significant impact on the total cost. Therefore, expanding the applicability of drones can effectively reduce the total cost.