Target Trajectory Prediction Based On Deep Neural Network

With the continuous development of science and technology,the situation of air combat is becoming more and more severe.The vigorous development of high-speed large maneuver targets has posed a great challenge to air combat maneuver decision making.Air combat trajectory prediction,as an important part of air combat maneuver decision,plays a pivotal role in air combat.The party with the ability to predict plays an irreplaceable role whether it is in a combat attack position or in a strategic defense position.Therefore,it is important to study the prediction of flight trajectory in air combat to guide the decision of aircraft to ensure the safety of our combatants.To address the importance of trajectory prediction in air warfare,the paper conducts a study on flight trajectory prediction for air combat based on deep neural networks,and the main work is as follows:(1)For the prediction problem of target 3D trajectory,this paper splits the 3D trajectory data into three sets of time series data of X-axis,Y-axis and Z-axis,and predicts the time series of the three axes respectively.Combined with Wavelet Decomposition(WD),we propose a flight trajectory prediction method that combines WD and Long Short-Term Memory(LSTM)network.The operation principle of this prediction model is to first decompose the trajectory data into three high-frequency components and one low-frequency component by wavelet decomposition,then predict the decomposed components separately,fuse the prediction results and compare the real trajectory to verify the accuracy of the prediction results.And compare other commonly used prediction models,and verify the effectiveness of the proposed model by error analysis.(2)In order to improve the prediction accuracy of the target trajectory and the generalization ability of the prediction model,for the spatial dimensional characteristics of the3 D trajectory data,this paper proposes the trajectory prediction based on Convolutional Neural Networks(CNN).The one-dimensional convolution can extract the spatial dimensional characteristics of the trajectory time series data,and the sparse connection and weight sharing of the CNN convolutional kernel are used to greatly reduce the number of weights,thus facilitating the optimization of the network and reducing the risk of overfitting.The validity of the model is verified through experiments and error analysis,and in order to verify the generality of the model,the CNN model is also used for multi-step prediction and comparison experiments with other models to verify that the model proposed in this paper has high prediction accuracy and provides a certain reference value for trajectory prediction.