关键词: 交通运输工程；停机位分配；优势演员-评论家算法；图模型；多目标优化；深度强化学习

Abstract: In air transportation, flights and aircraft parking positions at airports often change, and the efficient dynamic allocation of parking bays is necessary. An intelligent aircraft parking position allocation scheme based on dynamic task graphs and the advantage actor-critic algorithm (A2C) was proposed. A dynamic task graph model of flights and aircraft parking positions was constructed by the graph structure method. Graph feature vectors were extracted to form the task state space, and an intelligent agent model for aircraft parking position allocation was constructed. A solution method based on the advantage actor-critic algorithm was designed. Experimental research was conducted by the use of the data of Beijing Daxing International Airport. The results show that the proposed aircraft parking position allocation scheme increases the bridge-proximity rate of pre-allocated aircraft parking positions by 5.6%, and the bridge-proximity rate of dynamically allocated aircraft parking positions by 6.4%. The proposed allocation scheme enables efficient dynamic and intelligent allocation of aircraft parking positions and provides decision support for the scheduling of aircraft parking position resources at busy airports.

Key words: traffic and transportation engineering; gate assignment; advantage actor-critic algorithm; graphical model; multi-objective optimization; deep reinforcement learning