Title: A machine learning-based approach to predict random variation in the landing time of scheduled flights
Authors: Deepudev Sahadevan; Palanisamy Ponnusamy; Varun P. Gopi; Shivkumar Guruswami; Adithya K. Krishna
Addresses: Department of ECE, NIT Tiruchirappalli, Tiruchirappalli, Tamil Nadu, India ' Department of ECE, NIT Tiruchirappalli, Tiruchirappalli, Tamil Nadu, India ' Department of ECE, NIT Tiruchirappalli, Tiruchirappalli, Tamil Nadu, India ' R&D Technical Centre, Airports Authority of India, Hyderabad, Telangana, India ' Department of ECE, NIT Tiruchirappalli, Tiruchirappalli, Tamil Nadu, India
Abstract: In air traffic management (ATM), reliable and accurate prediction of a scheduled flight's landing time (LDT) is one of the most important aspects of achieving improved arrival and departure sequencing on the runway. In this paper, the robustness of applying a machine learning based model in ATM is analysed, i.e., to estimate the random variations in flying time and thereby improve the reliability of the prediction of scheduled flights' landing times. Two significant models are proposed and analysed, viz., an experience-based model which is derived from the historical 4D trajectory profile to estimate the initial flying time and a nonlinear regression-based machine learning method to predict the random variations in flying time. Compared to the random forest (RF) model of nonlinear regression modelling, the M5P model offers better landing time estimates and both models offer more than a 50% improvement in prediction error with a lesser number of attributes.
Keywords: air traffic management; ATM; air traffic flow management; ATFM; landing time; LDT; machine learning; prediction; scheduled flights.
International Journal of Sustainable Aviation, 2021 Vol.7 No.4, pp.293 - 318
Received: 31 Mar 2021
Accepted: 06 Jun 2021
Published online: 14 Dec 2021 *