Modelling and Solving the Single-Airport Slot Allocation Problem

Katsigiannis, Fotios A. and Zografos, Konstantinos and Lulli, Guglielmo (2022) Modelling and Solving the Single-Airport Slot Allocation Problem. PhD thesis, Lancaster University.

Text (2022katsigiannisphd)
2022katsigiannisphd.pdf - Published Version
Available under License Creative Commons Attribution.
Download (10MB)

Abstract

Currently, there are about 200 overly congested airports where airport capacity does not suffice to accommodate airline demand. These airports play a critical role in the global air transport system since they concern 40% of global passenger demand and act as a bottleneck for the entire air transport system. This imbalance between airport capacity and airline demand leads to excessive delays, as well as multi-billion economic, and huge environmental and societal costs. Concurrently, the implementation of airport capacity expansion projects requires time, space and is subject to significant resistance from local communities. As a short to medium-term response, Airport Slot Allocation (ASA) has been used as the main demand management mechanism. The main goal of this thesis is to improve ASA decision-making through the proposition of models and algorithms that provide enhanced ASA decision support. In doing so, this thesis is organised into three distinct chapters that shed light on the following questions (I–V), which remain untapped by the existing literature. In parentheses, we identify the chapters of this thesis that relate to each research question. I. How to improve the modelling of airline demand flexibility and the utility that each airline assigns to each available airport slot? (Chapters 2 and 4) II. How can one model the dynamic and endogenous adaptation of the airport’s landside and airside infrastructure to the characteristics of airline demand? (Chapter 2) III. How to consider operational delays in strategic ASA decision-making? (Chapter 3) IV. How to involve the pertinent stakeholders into the ASA decision-making process to select a commonly agreed schedule; and how can one reduce the inherent decision-complexity without compromising the quality and diversity of the schedules presented to the decision-makers? (Chapter 3) V. Given that the ASA process involves airlines (submitting requests for slots) and coordinators (assigning slots to requests based on a set of rules and priorities), how can one jointly consider the interactions between these two sides to improve ASA decision-making? (Chapter 4) With regards to research questions (I) and (II), the thesis proposes a Mixed Integer Programming (MIP) model that considers airlines’ timing flexibility (research question I) and constraints that enable the dynamic and endogenous allocation of the airport’s resources (research question II). The proposed modelling variant addresses several additional problem characteristics and policy rules, and considers multiple efficiency objectives, while integrating all constraints that may affect airport slot scheduling decisions, including the asynchronous use of the different airport resources (runway, aprons, passenger terminal) and the endogenous consideration of the capabilities of the airport’s infrastructure to adapt to the airline demand’s characteristics and the aircraft/flight type associated with each request. The proposed model is integrated into a two-stage solution approach that considers all primary and several secondary policy rules of ASA. New combinatorial results and valid tightening inequalities that facilitate the solution of the problem are proposed and implemented. An extension of the above MIP model that considers the trade-offs among schedule displacement, maximum displacement, and the number of displaced requests, is integrated into a multi-objective solution framework. The proposed framework holistically considers the preferences of all ASA stakeholder groups (research question IV) concerning multiple performance metrics and models the operational delays associated with each airport schedule (research question III). The delays of each schedule/solution are macroscopically estimated, and a subtractive clustering algorithm and a parameter tuning routine reduce the inherent decision complexity by pruning non-dominated solutions without compromising the representativeness of the alternatives offered to the decision-makers (research question IV). Following the determination of the representative set, the expected delay estimates of each schedule are further refined by considering the whole airfield’s operations, the landside, and the airside infrastructure. The representative schedules are ranked based on the preferences of all ASA stakeholder groups concerning each schedule’s displacement-related and operational-delay performance. Finally, in considering the interactions between airlines’ timing flexibility and utility, and the policy-based priorities assigned by the coordinator to each request (research question V), the thesis models the ASA problem as a two-sided matching game and provides guarantees on the stability of the proposed schedules. A Stable Airport Slot Allocation Model (SASAM) capitalises on the flexibility considerations introduced for addressing research question (I) through the exploitation of data submitted by the airlines during the ASA process and provides functions that proxy each request’s value considering both the airlines’ timing flexibility for each submitted request and the requests’ prioritisation by the coordinators when considering the policy rules defining the ASA process. The thesis argues on the compliance of the proposed functions with the primary regulatory requirements of the ASA process and demonstrates their applicability for different types of slot requests. SASAM guarantees stability through sets of inequalities that prune allocations blocking the formation of stable schedules. A multi-objective Deferred-Acceptance (DA) algorithm guaranteeing the stability of each generated schedule is developed. The algorithm can generate all stable non-dominated points by considering the trade-off between the spilled airline and passenger demand and maximum displacement. The work conducted in this thesis addresses several problem characteristics and sheds light on their implications for ASA decision-making, hence having the potential to improve ASA decision-making. Our findings suggest that the consideration of airlines’ timing flexibility (research question I) results in improved capacity utilisation and scheduling efficiency. The endogenous consideration of the ability of the airport’s infrastructure to adapt to the characteristics of airline demand (research question II) enables a more efficient representation of airport declared capacity that results in the scheduling of additional requests. The concurrent consideration of airlines’ timing flexibility and the endogenous adaptation of airport resources to airline demand achieves an improved alignment between the airport infrastructure and the characteristics of airline demand, ergo proposing schedules of improved efficiency. The modelling and evaluation of the peak operational delays associated with the different airport schedules (research question III) provides allows the study of the implications of strategic ASA decision-making for operations and quantifies the impact of the airport’s declared capacity on each schedule’s operational performance. In considering the preferences of the relevant ASA stakeholders (airlines, coordinators, airport, and air traffic authorities) concerning multiple operational and strategic ASA efficiency metrics (research question IV) the thesis assesses the impact of alternative preference considerations and indicates a commonly preferred schedule that balances the stakeholders’ preferences. The proposition of representative subsets of alternative schedules reduces decision-complexity without significantly compromising the quality of the alternatives offered to the decision-making process (research question IV). The modelling of the ASA as a two-sided matching game (research question V), results in stable schedules consisting of request-to-slot assignments that provide no incentive to airlines and coordinators to reject or alter the proposed timings. Furthermore, the proposition of stable schedules results in more intensive use of airport capacity, while simultaneously improving scheduling efficiency. The models and algorithms developed as part of this thesis are tested using airline requests and airport capacity data from coordinated airports. Computational results that are relevant to the context of the considered airport instances provide evidence on the potential improvements for the current ASA process and facilitate data-driven policy and decision-making. In particular, with regards to the alignment of airline demand with the capabilities of the airport’s infrastructure (questions I and II), computational results report improved slot allocation efficiency and airport capacity utilisation, which for the considered airport instance translate to improvements ranging between 5-24% for various schedule performance metrics. In reducing the difficulty associated with the assessment of multiple ASA solutions by the stakeholders (question IV), instance-specific results suggest reductions to the number of alternative schedules by 87%, while maintaining the quality of the solutions presented to the stakeholders above 70% (expressed in relation to the initially considered set of schedules). Meanwhile, computational results suggest that the concurrent consideration of ASA stakeholders’ preferences (research question IV) with regards to both operational (research question III) and strategic performance metrics leads to alternative airport slot scheduling solutions that inform on the trade-offs between the schedules’ operational and strategic performance and the stakeholders’ preferences. Concerning research question (V), the application of SASAM and the DA algorithm suggest improvements to the number of unaccommodated flights and passengers (13 and 40% improvements) at the expense of requests concerning fewer passengers and days of operations (increasing the number of rejected requests by 1.2% in relation to the total number of submitted requests). The research conducted in this thesis aids in the identification of limitations that should be addressed by future studies to further improve ASA decision-making. First, the thesis focuses on exact solution approaches that consider the landside and airside infrastructure of the airport and generate multiple schedules. The proposition of pre-processing techniques that identify the bottleneck of the airport’s capacity, i.e., landside and/or airside, can be used to reduce the size of the proposed formulations and improve the required computational times. Meanwhile, the development of multi-objective heuristic algorithms that consider several problem characteristics and generate multiple efficient schedules in reasonable computational times, could extend the capabilities of the models propositioned in this thesis and provide decision support for some of the world’s most congested airports. Furthermore, the thesis models and evaluates the operational implications of strategic airport slot scheduling decisions. The explicit consideration of operational delays as an objective in ASA optimisation models and algorithms is an issue that merits investigation since it may further improve the operational performance of the generated schedules. In accordance with current practice, the models proposed in this work have considered deterministic capacity parameters. Perhaps, future research could propose formulations that consider stochastic representations of airport declared capacity and improve strategic ASA decision-making through the anticipation of operational uncertainty and weather-induced capacity reductions. Finally, in modelling airlines’ utility for each submitted request and available time slot the thesis proposes time-dependent functions that utilise available data to approximate airlines’ scheduling preferences. Future studies wishing to improve the accuracy of the proposed functions could utilise commercial data sources that provide route-specific information; or in cases that such data is unavailable, employ data mining and machine learning methodologies to extract airlines’ time-dependent utility and preferences.