Special Seminar in Computing + Mathematical Sciences

Flight delays result in significant costs to passengers, airlines, and society as a whole. A major cause of these delays is the uncertainty in weather conditions, which pose a challenge to the planning and execution of efficient operations. In this talk, I will present two examples of how operational data can be used to model, analyze and control air traffic operations, by accounting for uncertainties. In the first part of the talk, I will describe the development and validation of a queuing model of airport departure processes, that can predict the taxi-out time and queuing delay distributions for each flight, in addition to surface congestion levels and departure throughput. The optimal control policy that minimizes expected congestion levels while maximizing airport throughput can then be determined using dynamic programming. The second part of the talk will present a new integer programming approach for solving very large-scale air traffic flow management problems in the presence of capacity uncertainties. The proposed stochastic optimization algorithm uses column generation to determine the optimal departure times, routes and cancellation decisions for all flights in the air traffic network, allowing for recourse as the scenarios develop. Using nation-scale examples from the United States, I will demonstrate a scalable, parallel implementation that is fast enough for real-time implementation.