Optimizing Aircraft Route Planning Based on Data-Driven and Physics-Informed Wind Field Predictions

Abstract

Upper-air wind fields play a crucial role in aircraft navigation, directly impacting flight safety and operational efficiency. In this study, we propose an advanced route planning framework that integrates wind field predictions derived from a neural network-based approach. Specifically, we leverage the PredRNN Sequence-to-Sequence algorithm to predict wind fields up to 10 h in advance. The model is trained on grid-based wind speed data at an altitude of approximately 5500 m, focusing on major airline routes over China. Our approach demonstrates superior accuracy in wind field forecasting when compared to other neural network architectures. To achieve route planning in dynamic wind environments, we employ the A* algorithm. The results demonstrate that the proposed method effectively identifies routes that approximate the ideal trajectory while successfully avoiding areas with drastic wind speed changes, thereby enhancing both the efficiency and safety of flight operations.