4D trajectory prediction and conflict detection in terminal areas based on an improved convolutional network.
Journal:
PloS one
PMID:
39951407
Abstract
At present, the passenger traffic volume of civil aviation is gradually increasing, and the scale of the airline network is gradually expanding. In order to optimize the air traffic service mode more safely and scientifically, the International Civil Aviation Organization (ICAO) has proposed a new concept of trajectory based operation (TBO). According to the basic principle and core structure of TBO, aiming at improving the accuracy of Four Dimensional (4D) Trajectory prediction and the reliability of multi-trajectory flight conflict detection, the trajectory prediction model is established by using convolutional neural networks-bidirectional gated recurrent unit (CNN-BiGRU), and the conflict evaluation of prediction results is realized by using trajectory distance detection function. The simulation experiment shows that the simulation experiment is carried out by introducing the real automatic dependent surveillance-broadcast (ADS-B) historical track data in the terminal area of the busy airport. The experimental results are compared with the experimental results of the single long short-term memory (LSTM) model and the gated recurrent unit (GRU) model in the same data set. The results show that the CNN-BiGRU trajectory prediction model is superior to the comparison model in many evaluation indexes, and the conflict detection results can be evaluated for the future 800 seconds interval of the two trajectories.