Aircraft trajectory prediction based on wavelet transform and state space model

Aircraft trajectory prediction, as one of the important links in air combat maneuvering decision-making, plays a vital role in the air battlefield. To achieve higher accuracy and longer prediction length of aircraft trajectory prediction, this paper proposes a trajectory prediction model WTTr-M (wavelet transfer and transformer-Mamba) based on wavelet transform and state space module. The model considers the frequency domain characteristics of the trajectory and uses wavelet transform to extract the time-frequency characteristics of the target trajectory for learning. Meanwhile, a prediction model based on the state space module is adopted for prediction. Through this module, the long-order trajectory data can be selectively propagated or forgotten, thereby enhancing the model's ability to predict longer trajectories. The experiment is verified by the self-generated battlefield simulation trajectory data and civil aviation trajectory data. The experimental results show that WTTr-M can achieve higher prediction accuracy in longer trajectory prediction tasks than other methods.