Reconstructing Flight Plans from Trajectories for Traffic Simulations

In Air Traffic Management (ATM), accurate flight trajectory prediction is crucial for evaluating operational concepts and assessing fuel consumption, environmental impact, and other performance metrics. This prediction typically relies on Flight Management Systems (FMS), which use flight plans as input and provide critical information on fuel burn, aircraft weight and other performance parameters. However, common air traffic data sources such as Automatic Dependent Surveillance-Broadcast (ADS-B) often only provide 4D waypoints and do not include the original filed flight plan. At the same time, the initial flight plan does not account for possible detours or controller interventions that would be needed for a complete analysis. This paper addresses these gaps by proposing an algorithm that reconstructs a suitable flight plan from a given trajectory, providing a complete representation of its path. To achieve this, the algorithm first generates a candidate flight plan based on predefined routes. Then, it compares the given trajectory to this candidate flight plan to identify possible detours, additional turns, or altitude constraints. The resulting flight plan combines these findings and provides a list of lateral waypoints that guide the aircraft from departure to arrival, as well as a cruise flight level estimate, altitude constraints, and additional information on turns. The algorithm was implemented as an additonal feature within the Future Air Traffic Simulator (FATS), an advanced simulation platform for the evaluation and validation of complex air traffic scenarios developed by the Institute of Flight Guidance at the German Aerospace Center (DLR), and validated using both synthetic and real-world data. The synthetic dataset consists of 862 short- and long-haul flights with introduced lateral and vertical deviations, while the real-world dataset includes a smaller number of ADS-B trajectories from the OpenSky Network. Using the trajectories as input, the algorithm successfully reconstructed suitable flight plans representing the entire flight path, including complex holding procedures. Accuracy is evaluated by comparing the original trajectory and reconstructed flight plan on a discretized grid. The results show a high level of agreement in the lateral dimension, whereas alignment in three dimensions is lower. This is mainly due to the current focus on level segments. Modelling the complete climb and descent procedures with altitude constraints will be incorporated in future work.