Student Name: Somto E. Okonkwo

Advisor: Dr. Dimitri Mavris

Milestone: PhD Thesis Proposal

Degree Program: Aerospace Engineering

Title: A Methodology for Dynamic Gate Allocation Using Deep Reinforcement Learning and Convective Weather Intelligence

Abstract: Since the first powered, sustained, and controlled flight by the Wright brothers in December 1903, the aviation industry has developed through rapid advancements in design and technology. This is seen in the safety, speed, comfort levels, and efficiency of air travel. As the world continues to feel more interconnected through aviation, there are operational and environmental costs that continue to be addressed to keep this industry profitable. This research is focused on the operational implications of air travel in the commercial sector. The overarching objective is to develop a dynamic gate allocation optimization framework using a deep reinforcement learning (DRL) approach under convective weather uncertainty, including a comparative evaluation against state-of-the-art learning-based algorithms across diverse airport operational environments. The first research area highlights weather as a predominant factor in airport operational delays and inefficiency. This is further compounded by the growing demand for air travel, particularly during the holiday months, which amplifies the vulnerability to convective weather disruptions. Existing models and decision-making tools lack integrated predictive weather intelligence and, instead, rely on empirical data. Among the numerous approaches to minimize flight delay, optimizing ground operations is common because of its dynamic and data-driven approach. However, gate allocation is the first bridge that connects passenger movement with aircraft serving. Thus, the broad objective of this research is to address the critical and underexplored area of dynamic gate allocation using advanced techniques and improving state-of-the-art methodologies. Data-driven models can be highly complex as the problem scales and even require large amounts of interactions if deep learning is applied. Methods for overcoming these limitations have not been extensively explored in the area of gate allocation problem (GAP). The second research area is motivated by the limited study on the potential of DRL algorithms to address ground delay through dynamic gate allocation. It incorporates real-time weather data into the decision-making state space and balances both passenger-oriented and airline-oriented objectives. Through feature ablation analysis, this research aims to show significant improvements in runtime efficiency and gate conflict reduction compared to state-of-the-art DRL algorithms.

Date and time: 2026-07-24, 9 am - 12 pm EST

Location: CoVE

Committee:
Dr. Dimitri Mavris (advisor), School of Aerospace Engineering
Dr. Kyriakos Vamvoudakis, School of Aerospace Engineering
Dr. Kai Wang, School of Computational Science and Engineering
Dr. Ameya Behere, School of Aerospace Engineering
Dr. Dushhyanth Rajaram, Zoox