HIGH DIMENSIONAL WEATHER DATA USED IN A DEEP GENERATIVE MODEL TO PREDICT TRAJECTORIES OF AIRCRAFT

Abstract


 
 
 The effectiveness of the aviation community depends on accurate forecasting of a 4D aircraft's trajectory, whether in real time or for counter-reality analysis. creating an effective tree-matching technique for the first time in this research to create feature maps that resemble images for historical flight trajectories using high- fidelity meteorological information, including wind, temperature, and convective conditions. Approach the orbit's tracking points as a conditional Gaussian mixture with parameters so they can benefit from our suggested integrated iterative neural network depth generation model. A network of mixed density LSTM decoders and a long memory (LSTM) encoder network make up the terminal. The decoder network learns additional spatial correlations-time from past flight routes and outputs the parameters of the Gaussian composite after the encoder network combines the most recent recorded flight plan information into fixed- length state variables. To learn feature representations from three-dimensional weather feature maps, transformation layers are added into the pipeline.