Data driven and learning based approaches for integrated landside airside operations optimization

• Main Author: Ali, Hasnain
• Other Authors: Sameer Alam
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Computer science and engineering; Science::Mathematics::Applied mathematics
• Online Access: https://hdl.handle.net/10356/160021
• Institution: Nanyang Technological University

Asia has witnessed a meteoric rise in its air traffic in the last decade; leading the region to experience the fastest air traffic growth, globally. However, air traffic infrastructure has not grown proportionately. Many Asian hubs are already operating beyond their designed capacity, resulting in delay proliferation at these airports. Current plans for airport capacity expansions---which are expensive, environmentally sensitive, and often infeasible---will fail to keep up with the predicted traffic demands. This is expected to cause congestion and lower efficiency of airport airside operations (e.g., higher taxi delays). Inefficient airside operations may cause snowball delay effects at terminal gates which, in turn, may impact the airport landside operations (e.g., transfer passenger connectivity). Therefore, airport planners and operators will be tasked with efficiently managing high traffic demands at airports. Airport taxi delays adversely affect airports and airlines around the world, resulting in airside congestion, increased Air Traffic Controllers workload, and adverse environmental impact due to excessive fuel burn. Existing approaches to reduce airside taxi delays include pushback control, efficient path planning, follow-the-greens, autonomous taxiing, etc. Airport Departure Metering (DM) or pushback control has garnered significant research interest in the last two decades. State-of-the-art DM methods use model-based DM policies that rely on airside departure modeling to obtain simplified analytical equations. Consequently, these models fail to capture non-stationarity in the airside operations causing DM policies to perform poorly under uncertainty---inherent to an airport environment. Besides, DM contains congestion by transferring taxiway delays to gates. Therefore, DM has the potential to disrupt scheduled gate assignments leading to passenger missed connections on landside. But current DM research addresses airside inefficiency in silo without considering its impact on landside passenger connectivity. The objective of this thesis is to develop novel methods and approaches, using airside surface movement data, for making airside operations efficient under uncertainty. Furthermore, this thesis proposes an integrated framework to investigate interactions between airside and landside operations for designing gate schedules which are sensitive to transfer journeys of passengers. The scope of the present study is limited to facilitating passenger connections (airline-to-airline transfer) while making airside movements efficient. There are three primary research contributions of this thesis. The first contribution develops a data-driven DM method to minimize airside surface congestion while maintaining runway throughput in an uncertain airport environment. The work casts the DM problem in a Markov Decision Process (MDP) framework and develops a representative airport-airside simulator, using historical surface movement data. The simulator incorporates uncertainties in taxiway movements and aircraft start times to introduce non-stationarity into the simulated environment for making the simulations realistic. The DM policy is, then, exposed to such simulated scenarios and tasked with recommending aircraft pushback times to minimize taxi delays while maintaining runway throughput. For effective state representation, this work introduces taxiway hotspot feature---regions where multiple aircraft may arrive concurrently---which improves the DM policy convergence rate during training. The performance of the learnt policy is evaluated under different traffic densities. Results, on a typical day of simulated operations at Singapore Changi Airport, demonstrate that DRL can learn an effective DM policy to contain congestion on the taxiways; reduce total fuel consumption by approx. 22%; and, altogether, better manage the airside traffic. The second contribution develops a passenger-centric model of airport operations to minimize missed connections in presence of arrival delays at an airport terminal serving low cost carriers. Such operations are highly sensitive to factors such as aircraft delays, turnaround time and flight connection time. The effects of these factors on self-connecting transfer passengers are carefully examined. Results show that the chances of missed connections can be significantly reduced by operationally maintaining higher turnaround time and minimum connection time and by bringing down delays at the airport. Specifically, by maintaining the flight turnaround time at 50 minutes, minimum connection time at 60 minutes and by containing arrival delays within 70% of the current delay spread at Singapore Changi Airport Terminal 4, transfer passenger missed connections can be prevented for almost all the flights. The third contribution proposes an integrated landside airside framework to investigate the interactions between landside and airside operations. The landside simulator simulates transfer passenger movements, and the airside simulator simulates taxiway movements based on historical surface movement data. Departing aircraft movements are then metered to reduce taxi and take-off delays under uncertainties. Delayed arrivals and/or departures may lead to conflicts at gate---rendering planned gate assignments infeasible or impractical. Using Singapore Changi Airport, as a case study, it is found that DM may transfer as much as 12 minutes of waiting time from taxiways to gates. This leads to increased gate conflicts which additionally delays the gate-in time of an arriving aircraft by 2-7 minutes on average. A minimum connection time of 70 minutes is found sufficient, in presence of DM induced delays, to reduce the probability of missed connections for transfer passengers. Congested airports---acting as bottleneck nodes in global aviation network---have far reaching effects on the entire air transportation system. To improve overall efficiency of future airport operations, there is an urgent need for holistically viewing airport as a complex system-of-systems and carefully examining the interdependent nature of airport operations. As copious amount of data becomes easily available, our ability to learn, draw insights and obtain actionable knowledge using past operations data is a pre-requisite for establishing efficient airport operations that are sensitive to passenger needs.