一种基于C-GRU飞行轨迹预测方法

飞行轨迹是由大量时间序列数据组成,遵循一定的运动规则。对敌方巡逻机飞行轨迹进行预测,能有效的提高战斗机生存率。针对现有单一预测模型对飞行轨迹预测精度不高,提出一种基于复合门控循环单元飞行轨迹预测方法。使用飞行仿真模拟获取多组飞行轨迹坐标点数据,用于复合门控循环单元网络模型参数训练,实现飞行轨迹预测。通过仿真结果分析,复合门控循环单元网络模型在X、Y、Z轴上的多组预测数据平均绝对误差在4.5 m内,且网络模型预测平均时间开销约为4.1 ms;使用平均绝对误差变化较大与较小的轨迹数据进行对比,其Y与Z轴的两组均方根误差相近。同时与门控循环单元、长短期记忆网络模型对比,其误差最小,在平均耗时接近的情况下预测的结果更加准确。所以本文提出的模型适用于不同的飞行轨迹,而且预测结果具有较高的可信度。

A C-GRU based flight trajectory prediction method

The flight trajectory is composed of a large amount of time series data and follows certain motion rules. Predicting the flight trajectory of enemy patrol aircraft can effectively improve the survival rate of fighter jets. This paper proposes a flight trajectory prediction method based on C-GRU, aiming at the low accuracy of flight trajectory prediction by the existing single prediction model. Use flight simulation to obtain multiple sets of flight trajectory coordinate point data for C-GRU network model parameter training to achieve flight trajectory prediction. Through the analysis of simulation results, the average absolute error of multiple sets of prediction data on the X, Y, and Z axes of the C-GRU network model is within 4.5m, and the average time overhead of network model prediction is about 4.1ms; Compared with the smaller trajectory data, the two sets of root mean square errors of the Y and Z axes are similar. At the same time, compared with the GRU and LSTM network models, the error is the smallest, and the predicted results are more accurate when the average time-consuming is close. Therefore, the model proposed in this paper is suitable for different flight trajectories, and the prediction results have high reliability.