基于学习的中段反导拦截时间和拦截点预测方法

弹道导弹实时、准确地预测拦截弹的拦截点与拦截时间,是实现中段突防的有效手段。针对弹道导弹中段突防中的拦截点坐标及拦截时间的预测问题,提出了一种基于监督学习的在线预测方法。以拦截弹的主动段关机参数和关机时刻为输入量,建立拦截时间和拦截点预测模型。在多层感知机神经网络的基础上构建有监督学习算法,通过攻防仿真获取拦截弹的参数制作训练数据集,在线下完成网络训练。仿真结果表明:神经网络能够有效在线预测拦截时间和拦截点坐标,预测结果的相对误差分别为0.124 3%和0.128 5%,拦截时间预测结果误差的平均值为0.224 0 s,拦截点预测结果距离误差平均值为2 016.48 m,均满足精度要求。 

Prediction method of intercept time and intercept point based on learning mid-course antimissile

Accurately predicting the intercept point and intercept time of the interceptor in real time is an effective way to realize the mid-course penetration of ballistic missiles. In order to predict the intercept point coordinates and intercept time during the mid-course penetration process of ballistic missile, an online prediction method based on supervised learning is proposed in this paper. Using the shutdown parameters and the shutdown time of the boost stage of the interceptor as inputs, the prediction model of intercept time and intercept point was established. Based on the multi-layer perceptron neural network, a supervised learning algorithm was formulated, and the interceptor's parameters were obtained through the attack and defense simulation to make the set of training data. The network training was completed offline. The simulation results show that the neural network can effectively predict the interception time and the coordinates of interception point online, and the relative error of the prediction results is 0.124 3% and 0.128 5% respectively; the average error of the prediction results of intercept time is 0.224 0 s; the average distance error of the prediction results of intercept point is 2 016.48 m. They all meet the accuracy requirements.