Skip to main content

Risk-Adjusted Traffic Management Strategies for Convective Weather Conditions

Paper ID

ATM-2023-026

Conference

USA/Europe ATM R&D Seminar

Year

2023

Theme

Weather in ATM

Project Name

Keywords:

airspace capacity, epsilon-greedy, reinforcement learning, risk, simulation, weather

Authors

James Jones and Zachary Ellenbogen

DOI

Project Number

Abstract

Weather is a significant source of disruption and uncertainty for air traffic. This unpredictability can present significant challenges to traffic managers when managing airport and airspace resources. The lack of data-driven decision support tools to advise stakeholders on how to best deal with weather impacts also represents a critical shortfall in enabling improved decision-making within air traffic management. In this paper, we present an epsilon-greedy approach that incorporates risk-adjusted objectives into recommendations of Traffic Management Initiative (TMI) parameters during uncertain weather conditions. The method attempts to achieve the best performance within the context of some of the worst-case weather outcomes. The method is compared to a standard epsilon greedy approach that attempts to maximize the expected value of an objective. The two approaches are evaluated using a parallel fast-time simulation framework over various weather scenarios. A set of TMIs at airports and airspace resources is applied and tested against seven case days in which the airspace capacity in the Northeast United States was affected by convective weather. The risk-adjusted method is generally able to achieve a higher number of operations with lower amounts of airborne holding in adverse weather conditions. The results suggest that the approach could potentially aid more risk-averse air traffic stakeholders by supporting their operational planning.