You are here: Home 8th Seminar Papers Application of Reinforcement Learning ...
Document Actions

Paper 79 -- Application of Reinforcement Learning Algorithms for Predicting Taxi-out Times

Accurate estimation of taxi-out time in the presence of uncertainties in the National Airspace System (NAS) is essential for the development of a more efficient air traffic management system. The dynamic nature of operations in the NAS indicates that traditional regression methods characterized by constant parameters would be inadequate to capture variations in taxi-out time across a day. In this paper, we describe how to build a taxi-out time estimation model using Reinforcement Learning, and identify factors that influence taxi-out time through the day. Taxi-out time predictions for a flight are made 15 minutes in advance of scheduled gate pushback time. Results from a case study of Detroit International Airport (DTW), Tampa International Airport (TPA), and John F. Kennedy International Airport (JFK) are presented and analyzed.
Theme: Airport Operations
Keywords: Reinforcement Learning, Taxi-out time estimation
Posted by: Poornima Balakrishna / Other authors: Rajesh Ganesan, Lance Sherry
Note: Unset Received On Jan 23, 2009

| Site Map | Accessibility | Contact |

Hosted and maintained by EUROCONTROL