4D Trajectory Prediction Method Based On Neural Network

In order to meet the challenges brought by the expansion of the route network and the shortage of airspace resources,the Traiectory Based Operation(TBO)has been taken as the core operation concept of the next generation Air Traffic Management system by the International Civil Aviation Organization(ICAO).TBO takes the 4D trajectory prediction of the aircraft as a reference,and the trajectory is shared in the ATC system to accurately manage and control the operation of the aircraft.Therefore,exploring the high-precision 4D trajectory prediction technology is the core problem to be solved urgently.Aiming at the high-precision 4D trajectory prediction problem,this paper proposes a novel 4D trajectory prediction method based on deep learning neural network.This method mainly mines the spatial-temporal features of a large amount of historical flight data to establish a neural network prediction model,so as to achieve high-precision prediction of the 4D trajectory.First,it analyzed and preprocessed ADS-B historical trajectory data,introduced ADS-B surveillance technology and the format of ADS-B trajectory data,and adopted cubic spline interpolation algorithm to supplement the missing points of flight data.Then,considering the plentiful spatial-temporal features of the trajectory,a 4D trajectory prediction model based on the CNN-LSTM combined neural network was proposed.This model combines a Convolutional Neural Network(CNN)with a Long Short-Term Memory(LSTM).The CNN module is used to extract the spatial dimensional features of the adjacent aera of the trajectory,the LSTM module is used to extract the dependency of trajectory time dimension,and based on the full fusion of spatial and temporal features,4D trajectory prediction is achieved with high accuracy.In order to better verify the performance of the proposed model,a single LSTM and BP model were designed for comparison.Finally,the real ADS-B historical trajectory data on a certain route was used for experimental verification.Experiments show that the CNN-LSTM combined neural network model proposed in this paper has better prediction effect than the single neural network model,and it can better reflect the actual flight path of the aircraft.