Prediction of Intercept Point and Impact Time for Hypersonic Aircraft in Terminal Phase

Accurately predicting the intercept point and the impact time for hypersonic aircraft can provide a strategic advantage on the battlefield, improve interception success rates, and reduce the threat posed by enemy missiles. Based on this, this paper proposes a method using neural networks for forecasting the intercept point and remain flight time of hypersonic aircraft. First, the movement models of the hypersonic aircraft and the interceptor are established, and then proportional navigation guidance commands are generated to compute the trajectory through numerical integration. Next, the prediction neural network incorporating long short-term memory (LSTM) and self-attention mechanisms is constructed and trained using trajectory samples. It takes the data from the past N timesteps as input and outputs the intercept point coordinates and remain flight time, fitting the complex nonlinear relationship between the two. Finally, the performance of the neural network on the test set is analyzed. The results show that the intercept point error decreases as the interceptor approaches the hypersonic aircraft and can achieve high accuracy in the end, while the predicted remain flight time fits the theoretical value well during the whole flight.