基于残差神经网络和LSTM的飞行轨迹预测

Flight Trajectory Prediction Based on Residual Neural Network and LSTM

针对飞行轨迹预测精度不足和误差大的问题,提出了将残差神经网络与LSTM结合起来的飞行轨迹预测算法。首先,利用坐标转换将经度、纬度和海拔高度数据转化为飞机在地面坐标坐标系下的位置坐标,再对坐标数据进行归一化处理;其次,通过残差神经网络来读取序列并自动学习内部特征,最后用LSTM来处理数据的时间序列信息。实验结果表明,该模型与LSTM、CNN+LSTM模型相比损失函数、均方根误差和平均绝对误差更小,效果更好,预测轨迹能够反映真实轨迹的航迹变化。

Aiming at the problem of insufficient accuracy and large error in flight trajectory prediction,a flight trajectory pre⁃diction algorithm combining residual neural network and LSTM is proposed.Firstly,the longitude,latitude and altitude data are converted into the position coordinates of the aircraft in the ground coordinate system by coordinate conversion,and then the coordi⁃nate data are normalized.Secondly,the residual neural network is used to read the sequence and automatically learn the internal characteristics.Finally,the LSTM is used to process the time series information of the data.The experimental results show that com⁃pared with LSTM and CNN+LSTM models,the loss function,root mean square error and average absolute error of this model are smaller,and the effect is better.The predicted trajectory can reflect the track changes of the real trajectory.