With the continuous increase in ground traffic density at large hub airports and the parallel operation of multiple aircraft in complex taxiway networks, the number of potential conflict points increases exponentially. Existing rule-based conflict detection methods suffer from high false alarm rates and insufficient real-time performance in dynamic environments. This paper proposes a spatiotemporal trajectory association algorithm based on graph neural networks (GNNs). First, the airport taxiway network is abstracted as a topological graph structure, with nodes representing taxiway intersections and parking positions, and edges representing taxiing paths. Second, a spatiotemporal graph convolutional module is constructed to capture the spatial dependence and temporal evolution characteristics of aircraft trajectories. Then, a multi-head attention mechanism is designed to dynamically associate the interaction behavior patterns of multiple aircraft. Finally, real-time prediction of conflict risks is achieved through end-to-end training. In a three-month validation study using real-world operational data from an international hub airport, the algorithm maintains a high collision detection accuracy of 93.5%-94.9%, with a median false alarm rate of 8.85% and a response time reduced to a minimum of 1.9 seconds, providing effective technical support for airport ground operations safety management.
Chao Zhao (Thu,) studied this question.