Junzi Sun, Ph.D.

My other sites

mode-s.org openap.dev

Data Science & Artificial Intelligence

Data science and AI are integral to modern aviation, offering solutions for complex challenges in safety, performance, and efficiency. The research landscape includes advanced methodologies such as augmented transfer learning for environmental monitoring and statistical models for trajectory predictions. These techniques leverage open data for validation, providing a robust framework that significantly impacts the aviation research.

Highlights

Whisper-ATC: Automatic Speech Recognition for Air Traffic Control
This study introduces the newly developed transformer model for air traffic control and provides a set of fully open automatic speech recognition models with high accuracies. The OpenAI Whisper models are fine-tuned with various air traffic control datasets. We evaluated the performance of different model sizes, and it was possible to achieve word error rates of 13.5% on the ATCO2 dataset and 1.17% on the ATCOSIM dataset with a random split (or 3.88% with speaker split).
Contrail-Net: Neural networks models for contrail detection and segmentation
This is an open-source project implements contrail segmentation neural network models in PyTorch. The models are built using augmented transfer learning, where I applied several image augmentation on a pre-train ResUNet model. This way, the model can be quickly fine tuned with a handful of labelled satellite images. With the invention of a new loss function, SR Loss, I can further optimizes the contrail detection using contrail information in Hough space.
Machine Learning Air Traffic Delays Prediction Models
In this open project, three Machine learning models are developed for predicting air traffic delays at different levels, which are flight-level, airport-level, and network-level. 1) Random Forest model to predict arrival delays for individual flights, 2) LSTM model to predict aggregated arrival and departure delays for a single airport, and 3) a dynamic spatial-temporal graph attention network model that predicts aggregated arrival and departure delays of all airports a network.

Related publications

  1. Automatic Control With Human-Like Reasoning: Exploring Language Model Embodied Air Traffic Agents
    Conference 2024
    Andriuškevičius, Justas ; and Sun, Junzi
    In Proceedings of the 14th SESAR Innovation Days, Rome, Italy
    Copy Andriuškevičius, J., & Sun, J. (2024, November). Automatic Control With Human-Like Reasoning: Exploring Language Model Embodied Air Traffic Agents. Proceedings of the 14th SESAR Innovation Days, Rome, Italy.
    Copy 
    @inproceedings{andriuskevicius2024automatic,
  title = {Automatic Control With Human-Like Reasoning: Exploring Language Model Embodied Air Traffic Agents},
  author = {Andriuškevičius, Justas and Sun, Junzi},
  booktitle = {Proceedings of the 14th SESAR Innovation Days, Rome, Italy},
  year = {2024},
  month = nov,
  link = {https://www.sesarju.eu/sites/default/files/documents/sid/2024/papers/SIDs_2024_paper_047%20final.pdf},
  tag = {ai}
}
  2. Whisper-ATC: Open Models for Air Traffic Control Automatic Speech Recognition with Accuracy
    Conference 2024
    Doorn, Jan ; Sun, Junzi ; Hoekstra, Jacco ; Jonk, Patrick ; and Vries, Vincent
    In Proceedings of the 11th International Conference for Research in Air Transportation
    Copy van Doorn, J., Sun, J., Hoekstra, J., Jonk, P., & de Vries, V. (2024, July). Whisper-ATC: Open Models for Air Traffic Control Automatic Speech Recognition with Accuracy. Proceedings of the 11th International Conference for Research in Air Transportation.
    Copy 
    @inproceedings{vandoorn2024whisperatc,
  title = {Whisper-ATC: Open Models for Air Traffic Control Automatic Speech Recognition with Accuracy},
  author = {van Doorn, Jan and Sun, Junzi and Hoekstra, Jacco and Jonk, Patrick and de Vries, Vincent},
  booktitle = {Proceedings of the 11th International Conference for Research in Air Transportation},
  year = {2024},
  month = jul,
  link = {https://resolver.tudelft.nl/uuid:8e02d222-5775-441d-94d2-96c26156cf43},
  tag = {ai}
}
  3. A Transformer-based Trajectory Prediction Model to Support Air Traffic Demand Forecasting
    Conference 2024
    Vos, Reinier ; Sun, Junzi ; and Hoekstra, Jacco
    In Proceedings of the 11th International Conference for Research in Air Transportation
    Copy Vos, R., Sun, J., & Hoekstra, J. (2024, July). A Transformer-based Trajectory Prediction Model to Support Air Traffic Demand Forecasting. Proceedings of the 11th International Conference for Research in Air Transportation.
    Copy 
    @inproceedings{vos2024transformer,
  title = {A Transformer-based Trajectory Prediction Model to Support Air Traffic Demand Forecasting},
  author = {Vos, Reinier and Sun, Junzi and Hoekstra, Jacco},
  booktitle = {Proceedings of the 11th International Conference for Research in Air Transportation},
  year = {2024},
  month = jul,
  link = {https://resolver.tudelft.nl/uuid:6deb30d1-b207-4fc7-adb5-07ca9bc29385},
  tag = {ai}
}
  4. Contrail Altitude Estimation Based on Shadows Detected in Landsat Imagery
    Conference 2023
    Roosenbrand, Esther ; Sun, Junzi ; and Hoesktra, Jacco
    In Proceedings of the 13th SESAR Innovation Days, Sevilla, Spain
    Copy Roosenbrand, E., Sun, J., & Hoesktra, J. (2023, December). Contrail Altitude Estimation Based on Shadows Detected in Landsat Imagery. Proceedings of the 13th SESAR Innovation Days, Sevilla, Spain.
    Copy 
    @inproceedings{roosenbrand2023altitude,
  title = {Contrail Altitude Estimation Based on Shadows Detected in Landsat Imagery},
  author = {Roosenbrand, Esther and Sun, Junzi and Hoesktra, Jacco},
  booktitle = {Proceedings of the 13th SESAR Innovation Days, Sevilla, Spain},
  year = {2023},
  month = dec,
  link = {https://doi.org/10.61009/SID.2023.1.06},
  tag = {sus, ai}
}
  5. Flight Contrail Segmentation via Augmented Transfer Learning with Novel SR Loss Function in Hough Space
    preprint 2023
    Sun, Junzi ; and Roosenbrand, Esther
    Copy Sun, J., & Roosenbrand, E. (2023). Flight Contrail Segmentation via Augmented Transfer Learning with Novel SR Loss Function in Hough Space.
    Copy 
    @preprint{sun2023flight,
  title = {Flight Contrail Segmentation via Augmented Transfer Learning with Novel SR Loss Function in Hough Space},
  author = {Sun, Junzi and Roosenbrand, Esther},
  year = {2023},
  eprint = {2307.12032},
  archiveprefix = {arXiv},
  primaryclass = {cs.CV},
  tag = {ai, sus}
}
  6. Designing Recurrent and Graph Neural Networks to Predict Airport and Air Traffic Network Delays
    Conference 2022
    Sun, Junzi ; Dijkstra, Tristan ; Aristodemou, Constantinos ; Buzetelu, Vlad ; Falat, Theo ; Hogenelst, Tim ; Prins, Niels ; and Slijper, Benjamin
    In Proceedings of the 10th International Conference for Research in Air Transportation
    Copy Sun, J., Dijkstra, T., Aristodemou, C., Buzetelu, V., Falat, T., Hogenelst, T., Prins, N., & Slijper, B. (2022, June). Designing Recurrent and Graph Neural Networks to Predict Airport and Air Traffic Network Delays. Proceedings of the 10th International Conference for Research in Air Transportation.
    Copy 
    @inproceedings{sun2022gatdelay,
  title = {Designing Recurrent and Graph Neural Networks to Predict Airport and Air Traffic Network Delays},
  author = {Sun, Junzi and Dijkstra, Tristan and Aristodemou, Constantinos and Buzetelu, Vlad and Falat, Theo and Hogenelst, Tim and Prins, Niels and Slijper, Benjamin},
  booktitle = {Proceedings of the 10th International Conference for Research in Air Transportation},
  year = {2022},
  month = jun,
  link = {https://research.tudelft.nl/files/126274370/ICRAT2022_paper_37.pdf},
  tag = {ai},
  award = true
}
  7. Quantifying Accuracy and Uncertainty in Data-Driven Flight Trajectory Predictions with Gaussian Process Regression
    Conference 2021
    Graas, Rik ; Sun, Junzi ; and Hoekstra, Jacco
    In Proceedings of the 11th SESAR Innovation Days, Online
    Copy Graas, R., Sun, J., & Hoekstra, J. (2021, December). Quantifying Accuracy and Uncertainty in Data-Driven Flight Trajectory Predictions with Gaussian Process Regression. Proceedings of the 11th SESAR Innovation Days, Online.
    Copy 
    @inproceedings{graas2021gprtp,
  title = {Quantifying Accuracy and Uncertainty in Data-Driven Flight Trajectory Predictions with Gaussian Process Regression},
  author = {Graas, Rik and Sun, Junzi and Hoekstra, Jacco},
  booktitle = {Proceedings of the 11th SESAR Innovation Days, Online},
  year = {2021},
  month = dec,
  tag = {ai},
  link = {https://www.sesarju.eu/sites/default/files/documents/sid/2021/papers/SIDs_2021_paper_70.pdf}
}
  8. A Framework to Evaluate Aircraft Trajectory Generation Methods
    Conference 2021
    Olive, Xavier ; Sun, Junzi ; Murça, M ; and Krauth, Timothé
    In 14th USA/Europe Air Traffic Management Research and Development Seminar
    Copy Olive, X., Sun, J., Murça, M., & Krauth, T. (2021, September). A Framework to Evaluate Aircraft Trajectory Generation Methods. 14th USA/Europe Air Traffic Management Research and Development Seminar.
    Copy 
    @inproceedings{olive2021framework,
  title = {A Framework to Evaluate Aircraft Trajectory Generation Methods},
  author = {Olive, Xavier and Sun, Junzi and Mur{\c{c}}a, M and Krauth, Timoth{\'e}},
  booktitle = {14th USA/Europe Air Traffic Management Research and Development Seminar},
  organization = {FAA/EUROCONTROL},
  year = {2021},
  month = sep,
  link = {https://research.tudelft.nl/files/105023386/ATM_Seminar_2021_paper_25.pdf},
  tag = {ai}
}
  9. Modeling and detecting anomalous safety events in approach flights using ADS-B data
    Conference 2021
    Bonifazi, Alberto ; Sun, Junzi ; Hoekstra, Jacco ; and Baren, Gerben
    In 14th USA/Europe Air Traffic Management Research and Development Seminar
    Copy Bonifazi, A., Sun, J., Hoekstra, J., & van Baren, G. (2021, September). Modeling and detecting anomalous safety events in approach flights using ADS-B data. 14th USA/Europe Air Traffic Management Research and Development Seminar.
    Copy 
    @inproceedings{sun2021safety,
  title = {Modeling and detecting anomalous safety events in approach flights using ADS-B data},
  author = {Bonifazi, Alberto and Sun, Junzi and Hoekstra, Jacco and van Baren, Gerben},
  booktitle = {14th USA/Europe Air Traffic Management Research and Development Seminar},
  organization = {FAA/EUROCONTROL},
  year = {2021},
  month = sep,
  tag = {ai},
  link = {https://research.tudelft.nl/files/105023158/ATM_Seminar_2021_paper_69.pdf}
}
  10. Estimating aircraft drag polar using open flight surveillance data and a stochastic total energy model
    Journal 2020
    DOI: 10.1016/j.trc.2020.01.026
    Sun, Junzi ; Hoekstra, Jacco ; and Ellerbroek, Joost
    Transportation Research Part C: Emerging Technologies
    Copy Sun, J., Hoekstra, J., & Ellerbroek, J. (2020). Estimating aircraft drag polar using open flight surveillance data and a stochastic total energy model. Transportation Research Part C: Emerging Technologies, 114, 391–404. https://doi.org/10.1016/j.trc.2020.01.026
    Copy 
    @article{sun2020estimating,
  title = {Estimating aircraft drag polar using open flight surveillance data and a stochastic total energy model},
  author = {Sun, Junzi and Hoekstra, Jacco and Ellerbroek, Joost},
  journal = {Transportation Research Part C: Emerging Technologies},
  volume = {114},
  pages = {391--404},
  year = {2020},
  month = may,
  publisher = {Elsevier},
  link = {https://research.tudelft.nl/files/71038050/published_OpenAP_drag_polar.pdf},
  doi = {10.1016/j.trc.2020.01.026},
  tag = {ai, perf}
}
  11. Aircraft initial mass estimation using Bayesian inference method
    Journal 2018
    DOI: 10.1016/j.trc.2018.02.022
    Sun, J. ; Ellerbroek, J. ; and Hoekstra, J.M.
    Transportation Research Part C: Emerging Technologies
    Copy Sun, J., Ellerbroek, J., & Hoekstra, J. M. (2018). Aircraft initial mass estimation using Bayesian inference method. Transportation Research Part C: Emerging Technologies, 90, 59–73. https://doi.org/10.1016/j.trc.2018.02.022
    Copy 
    @article{sun2018bayes,
  title = {Aircraft initial mass estimation using Bayesian inference method},
  author = {Sun, J. and Ellerbroek, J. and Hoekstra, J.M.},
  doi = {10.1016/j.trc.2018.02.022},
  journal = {Transportation Research Part C: Emerging Technologies},
  pages = {59--73},
  publisher = {Elsevier},
  volume = {90},
  year = {2018},
  month = may,
  tag = {ai},
  link = {https://research.tudelft.nl/files/52637982/1_s2.0_S0968090X18302626_main.pdf}
}
  12. Flight Extraction and Phase Identification for Large Automatic Dependent Surveillance–Broadcast Datasets
    Journal 2017
    DOI: 10.2514/1.I010520
    Sun, Junzi ; Ellerbroek, Joost ; and Hoekstra, Jacco
    Journal of Aerospace Information Systems
    Copy Sun, J., Ellerbroek, J., & Hoekstra, J. (2017). Flight Extraction and Phase Identification for Large Automatic Dependent Surveillance–Broadcast Datasets. Journal of Aerospace Information Systems, 14(10), 566–571. https://doi.org/10.2514/1.I010520
    Copy 
    @article{sun2017fdp,
  title = {Flight Extraction and Phase Identification for Large Automatic Dependent Surveillance–Broadcast Datasets},
  author = {Sun, Junzi and Ellerbroek, Joost and Hoekstra, Jacco},
  doi = {10.2514/1.I010520},
  journal = {Journal of Aerospace Information Systems},
  number = {10},
  pages = {566--571},
  publisher = {American Institute of Aeronautics and Astronautics},
  volume = {14},
  year = {2017},
  month = aug,
  tag = {ai},
  link = {https://research.tudelft.nl/files/25481708/main1.pdf}
}

© Copyright 2025 Junzi Sun