Aiming at the intricate optimization task of collaborative scheduling of multiple types of ground service vehicles in the ramp environment, a deep reinforcement learning algorithm integrated with the Transformer architecture is proposed in this paper. Firstly, in accordance with the varying priorities of the flight ground service process, the entire ground service task is decomposed, and subsequently, the complex multi-type vehicle scheduling issue is transformed into a sequential single-type vehicle scheduling problem. Then, the Transformer architecture is employed to automatically extract the features of flights and vehicles, and the task scheduling is resolved step by step through the decoder. The preliminary scheduling strategies are separately generated by combining the greedy algorithm and the Monte Carlo simulation algorithm, and these strategies are applied to the solution process of each sub-problem. On this basis, the deep reinforcement learning algorithm is utilized to train the entire model, and the scheduling strategy is continuously optimized through the interaction between the agent and the environment. Additionally, to enhance the robustness of the model and its capability to handle complex situations, this paper also trains the model by expanding the real data set. Finally, a considerable number of experiments have demonstrated that the deep reinforcement learning approach based on the Transformer architecture can effectively prevent mutual interference among different types of vehicles and capably cope with flight scheduling requirements in real environments.