Colloquium: Sebastiaan Camille Pepijn Wijnands (C&O)

The increasing complexity of air traffic management (ATM) systems calls for advancements in automation to ensure safety and efficiency. Artificial intelligence (AI) and machine learning (ML) offer powerful solutions, but their effectiveness depends on the availability of high-quality data, which is often limited. Synthetic data generation (SDG) provides a viable approach to address this limitation, enabling the creation of realistic datasets for ML training. This research focuses on applying TimeGAN to generate synthetic 4-dimensional aircraft landing trajectories in the Terminal Maneuvering Area (TMA), known for its high traffic density and trajectory diversity. The synthetic trajectories were assessed based on data diversity, fidelity, and usefulness. A significant challenge was imbalances in the training data, which impacted the model's ability to capture less frequent scenarios accurately. Generating synthetic data by grouping separate classes showed promise in mitigating these issues. TimeGAN demonstrated its ability to produce realistic trajectories that closely resemble historical data.

Supervisor: Dr. O.A. Sharpanskykh