A Data-Driven Framework for Air Traffic Network Analysis Using Aircraft Tracking Data


Student thesis: Doctoral Thesis

View graph of relations



Awarding Institution
Award date8 Apr 2019


Air transport system capacity enhancements cannot keep up with the increasing pace of demand growth in some part of the world, causing severe air traffic congestion and flight delays, which wastes time, fuel and takes a toll on economic activities and the environment. To tackle the issue, it is necessary to characterize the structure and dynamics of the air transport system, identify bottlenecks, and develop improvement strategies. However, to obtain a full picture of a national air traffic network represents a significant challenge due to the limited availability of operational data across different sources and regions. New opportunities have arisen from the increasing availability of digitized air traffic data, such as Automatic Dependent Surveillance-Broadcast (ADS-B) data, which makes it possible to track and analyze aircraft movement data on a global scale. In response, the thesis proposes a data-driven framework to 1) characterize the operational structure and dynamics of a national-level air traffic network, 2) build a stochastic and dynamic queueing network of air traffic congestions, and 3) perform case studies to demonstrate potential applications of the framework in air traffic management improvement using real-time aircraft tracking data and historical operational data. Compare with existing methods, this framework takes actual air traffic and operations into account and doesn’t require extensive input information. It can be used for air transport system characterization, flight delay prediction, and air traffic management (ATM) improvement strategies evaluation.

We first develop a novel data-driven framework that characterizes the operational structure and dynamics of a national air traffic network using only large-scale actual aircraft tracking data, assessing the air route availability, network structure, and utilization patterns of an air traffic network. Then we apply the proposed framework to analyze and compare the air traffic network of China with the one of the US using one month of historical aircraft tracking data. These results indicate that China faces a greater chance of en-route congestion when compared with the US.

Secondly, we build a stochastic and dynamic queueing network which takes en-route constraints into consideration to model flight delays and delay propagation through a national air transport system. The conceptual en-route “congestion points” represent capacity constraints that are caused by overcrowded airspace. They are identified through cluster analysis of the en-route structure. We build a multi-layer air traffic network consisting of airports, en-route “congestion points” and operational air routes via unsupervised learning of aircraft tracking data. Then, we develop a multi-layer air traffic network delay (MATND) model based on the identified multi-layer air traffic network. The MATND incorporates queueing engines to calculate flight delays at each airport and “congestion point”, and a delay propagation algorithm to capture delay propagation through the network.

In the third part of the thesis, we describe the detailed implementation of the MATND model and demonstrate the potential applications on policy-oriented ATM improvement strategies evaluation, such as how much more air traffic can current airspace infrastructure accommodate; how much delay can be reduced if more airspace/air routes are opened. A dataset that consists of 30 days’ aircraft tracking data and flight scheduling data is used to construct a multi-layer air traffic network consisting of the 56 busiest airports and 116 en-route “congestion points” in China. Results show that the MATND model performs well on delay prediction and bottlenecks identification and could be used as a fast tool to evaluate ATM improvement strategies.