Original scientific paper
https://doi.org/10.7307/ptt.v36i3.354
Air Traffic Scenario Evaluation Based on Metric Learning
Dong SUI
; College of Civil Aviation, Nanjing University of Aeronautics and Astronautics
Qian LI
; College of Civil Aviation, Nanjing University of Aeronautics and Astronautics
Tingting ZHOU
; College of Civil Aviation, Nanjing University of Aeronautics and Astronautics
Kechen LIU
; College of Civil Aviation, Nanjing University of Aeronautics and Astronautics
Abstract
Air traffic scenario evaluation can support the optimisation of traffic flow and airspace configuration to improve the safety of air traffic control. Since the air traffic scenario is influenced by the interaction of multiple factors, and real labelled data are lacking, the feature index selection and scenario evaluation are challenging endeavours. In this study, indicators were selected from three dimensions: airspace structure, traffic characteristics and meteorological conditions. The evaluation indicators were quantitatively screened according to information importance and overlap. Utilising the flow control and traffic flow information, the authors defined the free and saturated states of the state interval and developed a metric-based learning method to calibrate the state samples. A multilayer perceptron regression model was employed to establish the mapping relationship between the feature indicators and air traffic scenario. The evaluation accuracy of the sample set from three sectors in Shanghai exceeded 80%, which verified the effectiveness of the scenario evaluation model. This contribution holds practical significance in enhancing the safety of airspace operations.
Keywords
air traffic safety; air traffic scenario; evaluation indicator; metric learning
Hrčak ID:
318726
URI
Publication date:
20.6.2024.
Visits: 243 *