Enhanced Flight Planning by introducing stochastic trajectory data
Final Report Abstract
The project "Optimized flight planning by means of stochastic trajectory modeling" analyzes uncertainties when predicting aircraft positions and convert them into a uncertainty metric. Its causes rely on various disturbance variables, preferably weather forecast inaccuracies, estimated aircraft parameters (mass, thrust), unavoidable modeling errors. In the literature, these are usually treated statistically independent, which is not true due to e.g., aerodynamic interdependencies. We use Monte Carlo simulations to consider those. The goal is to develop a cause-and-effect model that can reproduce stochastic disturbance variables case-dependent and predict the resulting trajectory uncertainty over several hours robustly. This way, individual flights can take advantage in short term and traffic flows strategically. The model is validated using performance comparison to the optimal trajectory. As a result, significant potential for robust flight planning, and fuel saving can be demonstrated using the resulting Corridor of Optimization (CoO). The analysis focuses two areas, trajectory optimization under uncertainty and stochastic trajectory prediction (TP). First, a flight planning mechanism is developed using the department’s TOolchain for Multi-criteria Aircraft Trajectory Optimization (TOMATO) for an ensemble of weather forecasts of the Global Forecast System. The CoO is determined per flight, considering all TOMATO-multi-criteria-optimized trajectories per ensemble. The CoO serves to exploit these uncertainties for wind, temperature and pressure e.g., during climb for an efficient traffic flow management. In-flight, we can react optimally to updated weather information while remaining inside the CoO. Its appropriate dimensioning leads to robust and yet efficient TP. Second, a TP using Monte Carlo simulation is developed, considering physical flight characteristics as stochastic output of an artificial neural network. Validation is performed exemplarily for the climb phase with unspecified aircraft mass and airspeed. The findings show that the temporal interdependencies do not allow a blanket evaluation despite good results: To abstract the Monte Carlo simulation, a Bayesian network is used, combining graph theory with stochastics. Core functions are the estimation of the above inferences mass, airspeed and wind, as well as the propagation of temporal deviations from the flight plan. This way, e.g., the arrival time can robustly be predicted and, applied to several trajectories, a conflict detection function can be added.
Publications
- „Innovative Luftraumüberwachung mittels stochastischer Flugbahnmodellierung“, Ingenieurspiegel, S. 28–32, März 2018
T. Zeh, J. Rosenow, und H. Fricke
- „Re-optimizaton of 4D-Flight Trajectories During Flight Considering Forecast Uncertainties“, Deutscher Luft- und Raumfahrtkongress (DLRK), Friedrichshafen, 2018
M. Lindner, T. Zeh, und H. Fricke
- „Interdependent Uncertainty Handling in Trajectory Prediction“, Aerospace, Bd. 6, Nr. 2, S. 15, Feb. 2019
T. Zeh, J. Rosenow, und H. Fricke
(See online at https://doi.org/10.3390/aerospace6020015) - “In-Flight Aircraft Trajectory Optimization within Corridors Defined by Ensemble Weather Forecasts,” Aerospace, Bd. 7, Nr. 10, S. 144, Oct. 2020
M. Lindner, J. Rosenow, T. Zeh, und H. Fricke
(See online at https://doi.org/10.3390/aerospace7100144) - „Prediction of the Propagation of Trajectory Uncertainty for Climbing Aircraft“, 39th AIAA/IEEE Digital Avionics Systems Conference, San Antonio, TX, 2020
T. Zeh, J. Rosenow, R. Alligier, und H. Fricke
(See online at https://doi.org/10.1109/DASC50938.2020.9256711)
