Machine learning (ML) techniques are applied to build predictive models for aircraft pushback time predictions given historical data describing multimodal influences at and around an airport. A wide variety of influences were considered, including baggage handling data, security wait times, parking garage data, airport traffic video, and roadway transit times, to name a few. Predictive models for departure pushback time predictions (delays) were created for two classes of ML techniques: 1) regression techniques focused on predicting pushback delay times for either groups of flights or individual flights, and 2) classification techniques for predicting the pushback delay category (none, minor, major, or severe delays expected) for either groups or individual departing flights. In the worst case, one general ML model designed to predict delays for all departing flights does not perform well under all conditions; however, flight-specific ML models should be used instead, as they have superior accuracy. Hence, it is recommended that the overall departure delay capabilities supported by the multimodal ML approach and designed for specific airports include tailored sets of ML models for flights operating at these airports.
Machine Learning of Multimodal Influences on Airport Pushback Delays
01.05.2025
Aufsatz (Konferenz) , Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Optimizing Pushback Decisions to Valuate Airport Surface Surveillance Information
Online Contents | 2012
|Research on Aircraft Pushback Procedure in Busy Airport
TIBKAT | 2010
|ICAS2014_0032: OPTIMAL PUSHBACK TIME WITH EXISTING UNCERTAINTIES AT BUSY AIRPORT
British Library Conference Proceedings | 2014
|Addressing the Pushback Time Allocation Problem at Heathrow Airport
Online Contents | 2013
|