Critical flight trajectory identification via machine learning for large-scale trajectory management

In order to make a trade-off between the ever-increasing air traffic demand and the limited air traffic management capability, an efficient trajectory management is in urgent need. Considering that network-wide trajectories in the airspace need to be modified in the operation, the complexity of achieving the global optimum distribution of trajectories will increase greatly. Aiming to reduce the number of trajectories that need to be adjusted, previous papers introduced a few methods of identifying critical trajectories from a large-scale trajectories in the operation plan. However, there are still many challenges like unsatisfactory time complexity in O(n ² ) time due to huge data volume of flight trajectories. The main contribution of this paper is the development of a framework so that critical trajectories can be recognized more efficiently and accurately. In this framework, defined as the key feature set which for classification, several temporal-spatial operating characteristics of large-scale trajectories are summarized. After that, a machine learning model is used to classify the critical trajectories. As an exemplification, we develop a Support Vector Machine (SVM) model for its advantages in solving nonlinear and high-dimensional problems. The proposed method is demonstrated using real air traffic data collected from Chinese airspace. Results show that the method can improve the efficiency and accuracy of critical trajectories identification.