Multiobjective Optimisation of Aircraft Trajectories Under Wind Uncertainty Using GPU Parallelism and Genetic Algorithms

The future Air Traffic Management (ATM) system will feature trajectory-centric procedures that give airspace users greater flexibility in trajectory planning. However, uncertainty generates major challenges for the successful implementation of the future ATM paradigm, with meteorological uncertainty representing one of the most impactful sources. In this work, we address optimized flight planning taking into account wind uncertainty, which we model with meteorological Ensemble Prediction System forecasts. We develop and implement a Parallel Probabilistic Trajectory Prediction system on a GPGPU framework in order to simulate multiple flight plans under multiple meteorological scenarios in parallel. We then use it to solve multiobjective flight planning problems with the NSGA-II genetic algorithm, which we also partially parallelize. Results prove that the combined platform has high computational performance and is able to efficiently compute tradeoffs between fuel burn, flight duration and trajectory predictability within a few seconds, therefore constituting a useful tool for pre-tactical flight planning.