Indexado en

Licencia y uso

Citaciones

Altmetrics

Análisis de autorías institucional

Compartir

A NOVEL APPROACH FOR TURNAROUND TIME ALLOCATION BASED ON REINFORCEMENT LEARNING

Publicado en:32nd Congress Of The International Council Of The Aeronautical Sciences, Icas 2021. - 2021-01-01 (), DOI:

Autores: Rodríguez-Sanz Á; de la Cruz JH

Afiliaciones

Resumen

Airport ground processes are a critical path in the air traffic network. The management and reliability of the scheduled turnaround time is essential for the system's robustness and efficiency. This paper presents a method based on approximate dynamic programming and reinforcement learning to optimize airline policies regarding the scheduling of turnaround times. The model uses arrival delay and the airport level of congestion as state variables, while the turnaround time is the control variable. The feedback control approach is capable to adapt the proposed policy to changes in the system, which is a key feature due to the non-linear, stochastic and time-varying nature of airport operations. Moreover, the methodology provides a decision-making rule and not only a static optimal value. The cost function of the method considers delay and “buffer” costs, but also costs related to perturbations in schedule adherence, internal (local) delays and airport level of congestion. This approach widens the turnaround time allocation problem to a system-level view. The method is applied to a case study in a busy European airport, where the dataset covers approximately 34,000 turnaround operations. We focus the analysis on two carriers with different business models. Results show that an accurate use of turnaround time allocation methods may help airlines to manage the punctuality performance of turnaround aircraft by minimizing system costs.

Palabras clave