A complex network analysis approach for identifying air traffic congestion based on independent component analysis

Identifying air traffic congestion in air transportation domain is a challenging task. Most current algorithms are usually tested using some macro-indexes, such as aircraft density, aircraft clusters, stranded degree, and so on. In this work, we describe the air traffic situation from the perspective of complex networks: aircraft in airspace are regarded as nodes and edges form within airborne collision avoidance system (ACAS) communication ranges. The topological property indicators in complex networks, such as loop numbers, node strength, average clustering coefficient, between-ness centrality and weighted network efficiency, are adopted for dynamic air traffic situation. On this basis, several assessment methods can identify air traffic congestion. Nevertheless, since these methods need to set thresholds subjectively, we propose the use of independent component analysis (ICA) for online monitoring air traffic congestion. In this way, smooth situations are treated as datasets for training, and congested situations can be identify according to the change of SPE-statistic, l(2)-statistic and l(e)(2)-statistic. The simulation results show that proposed method has the ability to identify air traffic congestion well. (C) 2019 Published by Elsevier B.V.