Trajectory Prediction Algorithm of Missile Based on Physics-Augmented Deep Learning

Recently, high-precision trajectory prediction of ballistic missile in the boost phase has become a research hotspot in missile defense system. This article proposes a trajectory prediction algorithm based on physics-augmented deep learning (PADL) to solve this problem. First, the motion model of the boost phase is given. Then, the PADL trajectory prediction model is derived to accurately describe the complex dynamic characteristics of the boost phase. The model combines the advantages of physical model and deep-learning (DL) model, which enhances the model interpretability and reduces the model uncertainty. On this basis, this article designs an SECG DL network to deeply mine the time-series features in the target state data and predict the turning coefficient. The SECG network combines the convolutional neural network and the gated recurrent unit neural network, and introduces the squeeze-and-excitation networks. Finally, simulation verification is carried out under various conditions. The results show that the PADL trajectory prediction algorithm has high precision, good stability, and strong robustness, which can realize high-precision trajectory prediction of ballistic missile in the boost phase.