面向航班延误场景的机位预分配模型及算法研究

针对航班延误场景下易出现机位变更的问题,文章以最小化机位冲突概率和最大化乘客靠桥率为目标,增加基于机位冲突概率的鲁棒性约束,结合机场实际业务规则构建具有良好抗延误特性的机位预分配模型,并将其建模为马尔可夫决策模型,提出基于异步优势动作评价的机位预分配算法(gate assignment algorithm based on asynchronous advantage actor-critic, GABA3C)求解该问题。为验证所提算法在各种变化场景下的适用性,文章设置3组场景实例。仿真结果表明,所提出的算法在有效提升旅客满意度的同时,还可以解决因航班延误造成的机位冲突问题。相比于自适应并行遗传算法(adaptive parallel genetic algorithm, APGA)、近端策略优化(proximal policy optimization, PPO)算法以及深度Q网络(deep Q-network, DQN)算法,该文所提算法求得的解在乘客靠桥率上的目标值分别提高了5.7%、4.6%、5.8%,在机位冲突概率上的目标值分别降低了23.5%、10.0%、17.4%。

In order to solve the problem of gate change in the flight delay scenario, this paper aims to minimize the probability of gate conflict and maximize the near-gate passenger allocation rate, increases the robustness constraints based on the gate conflict probability, and constructs a gate pre-assignment model with good anti-delay characteristics combined with the actual airport business rules. It is modeled as a Markov decision model, and a gate assignment algorithm based on asynchronous advantage actor-critic(GABA3C) is proposed to solve this problem. In order to verify the applicability of the proposed algorithm in various changing scenarios, three sets of scenario examples are set up. Simulation results show that the proposed algorithm can effectively improve passenger satisfaction and solve the problem of gate conflict caused by flight delays. Compared with adaptive parallel genetic algorithm(APGA), proximal policy optimization(PPO) and deep Q-network(DQN) algorithm, the target value of solutions obtained by the proposed algorithm in the near-gate passenger allocation rate is increased by 5.7%, 4.6% and 5.8%, respectively, and the target value in the gate conflict probability is reduced by 23.5%, 10.0% and 17.4%, respectively.