使用神经网络的末制导方法

Neural Network Terminal Guidance

选择具有常值加速度的目标作为一个简化但有代表性的目标模型，以拦截弹的Δｖ最小为性能指标，得到了一定初始拦截距离下的训练数据，并使用ＢＰ算法对神经网络进行了训练。在有目标机动加速度估计误差的情况下，将神经网络制导获得的拦截精度和变轨机动速度增量同用扩展比例导引得到的结果进行了比较。结果表明，神经网络导引法能够保证对目标的拦截精度，具有很强的适应性。

In this paper, a 3 layers feed forward neural network trained by BP algorithm is selected and used in the terminal guidance. A target that has constant acceleration is selected as the target model that is simplified but representative. To minimize index performance of Δ v ,we can get training data of some intercept distance to train the neural network,then compare the intercept precision and the amount of Δ v between neural network guidance(NNG)and augmented proportional navigation (APN) under the condition of having target acceleration estimation error.