A fusion framework to characterize and evaluate air traffic clusters based on potential field theory

Abstract

The dramatic increase in flight volume and density has inevitably led to the small-scale clustering of aircraft, presenting significant challenges for air traffic controllers due to the complex interrelationships involved. In order to better characterize the clustering phenomenon and its complexity with massive real-time data, this paper introduces a data-driven fusion framework. This framework incorporates an interdependent network and a novel indicator system based on potential field theory to characterize the interaction and superposition effects within these clusters, encompassing aircraft, airspace waypoints, and air traffic controllers. To further explore the complexity of clusters according to the characteristic indicator system, a deep learning model is employed to identify these cluster scenarios. Training results on these scenarios indicate that the deep learning model, using the GRU-attention model, excels in capturing the complex features of clusters. In practical scenario experiments, the proposed fusion framework integrates the interdependent network, characteristic indicator system, and the GRU-attention model. It has proven effective in characterizing clusters and evaluating their complexities. This not only provides real-time alerts but also supports decision-making for air traffic controllers in challenging situations.

Keywords

• Air traffic complexity
• Interdependent network
• Potential field theory
• Deep learning