基于期望Sarsa的进港航班排序模型研究

Research on Arrival Flights Scheduling Model Based on Expected Sarsa

日益增长的航班流量给机场的维护带来了巨大压力的同时,也极大地影响乘客的体验。为了缓解机场的拥堵状况,结合人工智能算法,提出采用强化学习模型对终端区进港航班的序列进行优化。航班排序采用强化学习中的期望Sarsa算法,针对航班的延误考虑设计延误时间、延误成本、尾流类型等因素的奖励函数,以航班的预计达到时刻为状态,延误的分钟数为动作,对成都双流国际机场连续进港的20架航班进行仿真实验,结果表明优于现有的基于优先级的先到先服务算法和常见的启发式算法。

The increasing flow of flights has put greatly affecting on passenger’s experience while enormous pressure on airport maintenance.In order to alleviate the congestion of the airport,combined with artificial intelligence algorithms,it is proposed to use a reinforcement learning mod⁃el to optimize the flights scheduling in the terminal area.This method adopts the Expected-Sarsa algorithm in reinforcement learning.For flight delays,the reward function is designed with factors such as delay time,delay cost,wake type and so on.The estimated arrival time of the flight is the state and the action is the number of minutes of delay.The simulation experiment is carried out on 20 consecutive arriving flights of Chengdu Shuangliu International Airport,and the results show that it is superior to the existing priority-based FCFS algorithms and common heuristic algorithms.