Airspace conflict resolution is critical for safe operations of aircraft, becoming more important with ever-increasing airspace congestion. While there are numerous aircraft's conflict resolution approaches in the literature, almost all of them are based on flight dynamics to predict aircraft's future trajectories and generate a conflict resolution strategy by maneuvering and thus modifying flight paths. However, it is unclear how to analyze the current-day operations provided by air traffic controllers from the flight dynamics viewpoint. In this paper, we propose a data-driven resolution generator (D2RG) for air traffic control using machine learning, which guarantees safety. In the D2RG, a resolution strategy for a given conflict situation can be automatically synthesized based on the knowledge about the types and characteristics (or parameters) of resolutions managed by air traffic controllers, which is extracted from flight data. The proposed methodology is demonstrated with flight data from a multi-fidelity modeling and simulation system, and also tested with actual flight data to show its applicability to real scenarios.

解决空域冲突对于飞机的安全运行至关重要，随着空域拥挤程度的日益增加，这一点变得越来越重要。尽管文献中有许多飞机的冲突解决方法，但几乎所有方法都是基于飞行动力学来预测飞机的未来轨迹，并通过操纵和修改飞行路径来生成冲突解决策略。但是，目前尚不清楚如何从飞行动力学的角度分析空中交通管制员提供的当前运行情况。在本文中，我们提出了一种使用机器学习来控制空中交通的数据驱动分辨率生成器（D2RG），它可以确保安全性。在D2RG中，可以基于关于空中交通管制员管理的分辨率的类型和特征（或参数）的知识，自动从飞行数据中提取出针对给定冲突情况的解决策略。所提出的方法论通过多保真度建模和仿真系统的飞行数据进行了演示，并且还通过实际飞行数据进行了测试，以显示其在实际场景中的适用性。