Sequential Congestion Management with Weather Forecast Uncertainty

Abstract

En route airspace congestion, often due to convective weather, causes system-wide delays and disruption in the U.S. National Airspace System (NAS). Present-day methods for managing congestion are mostly manual, based on uncertain forecasts of weather and traffic demand, and often involve rerouting or delaying entire flows of aircraft. A sequential decision-making approach is proposed, in which traffic and weather forecast prediction uncertainty is quantified and explicitly used to develop efficient congestion resolution actions. The method is based on Monte Carlo simulation of traffic and weather outcomes from a specific forecast. Candidate sequential decision strategies are evaluated against the range of outcomes to determine the best course of action. Decisions are made based on a quantitative evaluation of the expected delay cost distribution, and resolution actions are targeted at specific flights, rather than flows. A weather-induced airspace congestion scenario is explored using the simulation, and three different levels of weather forecast uncertainty are postulated. Weather forecast uncertainty was shown to affect when and how aggressively to act to solve the congestion problem. Forecast uncertainty also increases the mean and variance of the congestion resolution cost. The simulation can be used both to learn about solving airspace congestion problems, and to do several types of cost benefit analyses. It is also a prototype of a future, real-time, probabilistic decision support aid for tactical traffic management.