Transformer based Classification of Ballistic Missiles from Radar Observations

초록

In this paper, the application of a transformer deep learning model is proposed for the classification of ballistic missile types using radar measurements-distance, angle, range rate, received power, and RCS-across various ranges. First, a scenario for missile trajectory generation is defined, and trajectory simulations are conducted for each missile type to examine differences in their radar measurements. Based on these simulations, a training dataset is constructed and a transformer-encoder-based model is developed. The learning performance is then evaluated with respect to the number of trajectories used and different combinations of input variables, and classification results are analyzed for cases in which distinct missile types exhibit similar trajectories.

키워드