TU Delft / AE / CO / CNS-ATM / Master thesis assignment:

Existing trajectory prediction studies in the literature often focus on getting the prediction right. New operation concepts, such as 4D TBO, rely heavily on the accurate prediction of the flights. However, due to the existence of many random factors, such as weather, flight procedures, congestion, and regulations, the prediction of flight is not always certain.

In this master thesis project, you will design different methods to tackle the unpredictability of flight trajectories caused by known and unknown random factors. You will make use of stochastic methods to measure and model the uncertainty in flight predictions, for example, by employing the combination of model-based and data-driven approaches. The goal is to model uncertainty caused by these random variables at different phases of flights, as well as to relate them to uncertainty spatial-temporal uncertainties of the aircraft.

This thesis will make use of several stochastic tools that we have proposed in our early research, such as hierarchical Bayesian computing, Gaussian Process Regression, and particle filtering.

During this master research project, you should be able to:

- Explore the trajectory data and apply the point-mass aircraft performance model.
- Aggregate trajectory data with additional datasets.
- Analyze large quantities of the flight trajectory data using customized tools developed in Python.
- Identify flight phases (segments) that are significantly challenging for prediction.
- Compare different data-driven and model-based trajectory prediction methods, and select the best method for the use case defined in this project.
- Identify external factors that influence the flight trajectories, and quantify their influences.
- Create datasets for training, test, and validation of the models.
- Construct stochastic prediction models (such as Gaussian Process Regression) for flight at different phases.
- Formulate data-driven strategies to model the uncertainty in predictions and causal relationships with different influence factors.
- Design methods to validate the models and algorithms proposed by the research.

In addition to related aeronautics background knowledge, we are looking for an MSc student who:

- has taken the AE4321 Air traffic management course as elective;
- is interested in data analysis and stochastic (Bayesian) computing;
- has a high level of programming skills (Python);
- can think and manage the progress independent;
- have a strong sense of time management.

- Dr. Junzi Sun [
*Daily supervisor*] - Prof.dr.ir Jacco Hoekstra

- Rudnyk, J., Ellerbroek, J. and Hoekstra, J.M., 2019. Trajectory Prediction Sensitivity Analysis Using Monte Carlo Simulations Based on Inputs’ Distributions. Journal of Air Transportation, 27(4), pp.181-198.
- Sun, J., Hoekstra, J.M. and 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, pp.391-404.
- Chati, Y.S. and Balakrishnan, H., 2018. Modeling of aircraft takeoff weight using gaussian processes. Journal of Air Transportation, 26(2), pp.70-79.
- Casado, E., La Civita, M., Vilaplana, M. and McGookin, E.W., 2017, September. Quantification of aircraft trajectory prediction uncertainty using polynomial chaos expansions. In 2017 IEEE/AIAA 36th Digital Avionics Systems Conference (DASC) (pp. 1-11). IEEE.
- Alligier, R., 2020. Predictive Distribution of Mass and Speed Profile to Improve Aircraft Climb Prediction. Journal of Air Transportation, pp.1-10.
- Alligier, R., Gianazza, D. and Durand, N., 2015. Machine learning and mass estimation methods for ground-based aircraft climb prediction. IEEE Transactions on Intelligent Transportation Systems, 16(6), pp.3138-3149.
- Sun, J., Vû, H., Ellerbroek, J. and Hoekstra, J.M., 2019. pymodes: Decoding mode-s surveillance data for open air transportation research. IEEE Transactions on Intelligent Transportation Systems.

© Copyright 2020 Junzi Sun