Prediction of Ballistic trajectories based on Gaussian Mixture Model
Jihuan Ren, Yi Liu, Xiang Wu, Yuming Bo · 2021
Existing trajectory prediction methods have some problems like low accuracy and poor real-time performance. The ballistic trajectory sampled by radar is essentially a continuous sequence, and the Gaussian Mixture Model (GMM) performs well in time-series prediction. To predict the trajectory more accurately, we construct a GMM with two different kernel functions weighted together. We build datasets of exterior trajectories under different initial conditions and train a GMM with optimal hyperparameters. Experimental results show that the GMM has higher prediction accuracy in the short term with about three times faster speed than the traditional Ballistic Differential Equations(BDE) method.