CEAS EuroGNC 2024
|
Contrastive Learning-Based Air Traffic Trajectory Representation: A Case Study on Incheon International Airport
|
| Thaweerath Phisannupawong |
M.S. student, Korea Advanced Institute of Science and Technology, Department of Aerospace Engineering, 34141, Daejeon, South Korea. | | Joshua Julian Damanik |
Ph.D. student, Korea Advanced Institute of Science and Technology, Department of Aerospace Engineering, 34141, Daejeon, South Korea. | | Han-Lim Choi |
Professor, Korea Advanced Institute of Science and Technology, Department of Aerospace Engineering, 34141, Daejeon, South Korea. |
|
| Abstract:
Air traffic trajectory recognition has become an interest in response to the airspace modernization. Similar to time series data, trajectory can be analyzed using representation learning. However, research on trajectory is less explored compared to other time series data. This paper introduces a machine learning-based approach to learning trajectory representations, which enhances performance in downstream recognition tasks. This contrastive representation learning framework is demonstrated on public unlabeled air traffic surveillance data. Using the contrastive objective, the model learns to maximize the agreement in representation for similar subseries, defined by the unchanged track, while distinguishing them from globally sampled negatives. The model uses sliding window encoding to transform the trajectories into more generalizable semantic terms, resulting in scalability for incomplete trajectories. The clusterability of the embeddings was compared with the clustering of the corresponding raw trajectories. The results suggest that analysis using the learned representation generates more elaborative clusters from a comprehensive point of view for both arrival and departure air traffic data.
|
| Keywords: Contrastive Learning; Air Traffic Management; Representation Learning; Time Series Analysis; Trajectory Clustering |
 View PDF CEAS-GNC-2024-063 |
| Thaweerath Phisannupawong, Joshua Julian Damanik, Han-Lim Choi: Contrastive Learning-Based Air Traffic Trajectory Representation: A Case Study on Incheon International Airport. Proceedings of the 2024 CEAS EuroGNC conference. Bristol, UK. June 2024. CEAS-GNC-2024-063.
|
| BibTeX entry:
@Incollection{CEAS-GNC-2024-063,
author = {Phisannupawong, Thaweerath and Damanik, Joshua Julian and Choi, Han-Lim},
title = {Contrastive Learning-Based Air Traffic Trajectory Representation: A Case Study on Incheon International Airport},
booktitle = {Proceedings of the 2024 {CEAS EuroGNC} conference},
address = {Bristol, UK},
month = jun,
year = {2024},
note = {CEAS-GNC-2024-063}
}
|