Zhang, S., Zhang, Y., Tay, T., & Shankar, J. (2022). Learning-based Aircraft Trajectory Analysis Tool for Holding and Vectoring Identification with ADS-B Data. 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC). https://doi.org/10.1109/itsc55140.2022.9921823
The post-analysis of actual flight trajectories is a crucial source for analyzing the efficiency of the air traffic management (ATM) and the workloads of air traffic controllers (ATCOs). The number of flights in holding and vectoring patterns is one aspect reflecting the workload and efficiency of the ATCOs and the capacity utilization of the airport. Here we propose a learning-based aircraft trajectory analysis tool for holding and vectoring event identification. This tool was designed with a small-scale Convolutional Neural Network (CNN) to ensure efficiency. The trajectory and Standard Instrument Departure Routes (SID)/Standard Arrival Routes (STAR) are represented in 2D-array form with separated channels. The proposed tool can be interpreted as an efficiency analyzer for trajectory planning or an indicator to assess TMA utilization. The proposal provides a fast evaluation tool capable of analyzing more than 100 trajectories every second with just a single laptop CPU and can be accelerated with GPU. The accuracy of identification is over 95% on the training set and over 90% on the test set. The training of the network requires the labeling of just 500 to 1000 trajectories (can be completed in hours by a single annotator). The proposal was also implemented to unlabeled datasets.
This research / project is supported by the National Research Foundation, Singapore - Aviation Transformation Programme
Grant Reference no. : ATP_IOP for ATM_I2R_2