A LSTM-Based Trajectory Prediction Method For Aerial Target

The dynamic and accurate prediction of the target trajectory about UAVs is the key point of airspace management. For the task of air target trajectory prediction dominated by a variety of UAVs, traditional methods can not deal with conditions such as complex modeling, high computational cost, and poor model reusability. This paper proposes an air target trajectory prediction method based on sliding window Long Short-Term Memory (LSTM). The sliding window method is adopted to segment the target historical trajectory data. The LSTM can memorize historical information and effectively process long time series to predict the trajectory data of air targets at future moments. The prediction error comparison simulation experiment is carried out between the polynomial fitting model method and the method proposed, which proves that the proposed method improves the trajectory prediction accuracy.