Application of Trajectory Clustering for Aircraft Conflict Detection

This paper presents the application of machine learning and open-source data to improve the prediction capability of conflict between aircraft in terminal airspace. An analytical methodology is developed that synthesizes trajectory clustering, classification-based supervised learning, and probabilistic modeling by leveraging publicly available flight data to streamline the probability of conflict and to facilitate the early detection of potential threat in terminal airspace. As such, application of this methodology yields an early detection of potential threats and allows for their resolution ahead of time. The methodology developed outlines a robust and repeatable framework that is applicable across heterogeneous data sets containing multiple aircraft and airport of operations. This framework is implemented on landing data at Zurich airport. The data are retrieved from the OpenSky Network which provides air traffic surveillance data from ADS-B messages. The work concludes by demonstrating the capability of the algorithm on a use case and provides initial guidance for its integration with existing collision avoidance systems.