一种基于反向神经网络的航空器飞行轨迹预测

为了缓解终端区空域拥堵和降低航空器运行风险,提出一种基于反向神经网络(BP)的航空器飞行轨迹预测模型。首先,对航空器历史数据进行筛选和降噪处理,得到基准轨迹;其次,建立基于Hausdorff距离的轨迹相似性矩阵,采用模糊C-均值聚类(FCM)对所有轨迹进行自动分类;最后,综合考虑飞行轨迹的三维位置、速度和航向特征,利用BP神经网络对轨迹特征进行训练学习,建立飞行轨迹预测模型,用于对未来时刻的短期飞行轨迹多维特征进行预测。试验结果表明:该网络模型预测误差小、预测效果好,可以更加准确地进行航空器的飞行轨迹预测。

Flight Trajectory Prediction of Aircraft Based on Back-propagation Neural Network

In order to alleviate the airspace congestion in Terminal Area and reduce the operational risk of aircraft, a flight trajectory prediction model based on back-propagation neural network (BP) was proposed. Firstly, the aircraft history data were filtered and denoised to obtain the reference trajectory; secondly, the trajectory similarity matrix based on Hausdorff distance was established, and all trajectories were classified automatically by using fuzzy C-means clustering (FCM); finally, considering the three-dimensional position, velocity and heading characteristics of flight trajectory, BP neural network was used to train and learn the trajectory characteristics. A flight trajectory prediction model was established to predict the multidimensional characteristics of the flight trajectory at the future time. The experimental results show that the prediction error of the network model is smaller, the prediction effect is better, and the flight trajectory prediction of aircraft can be more accurate.