LSTM-Based UAV Swarm Trajectory Prediction and Intent Recognition

This study focuses on the problem of trajectory prediction and intention recognition of UAV clusters in military defence, and explores a new method for UAV cluster trajectory prediction by constructing UAV cluster dataset autonomously and applying Long Short-Term Memory (LSTM) network for experimental validation. Traditional studies have mostly focused on the trajectory analysis of individual UAVs, while the dynamic behavioural patterns of UAV clusters are less studied. To this end, we generate a UAV cluster dataset covering multiple flight modes, and model and predict their trajectories using the LSTM model. The experimental results show that LSTM can effectively capture the temporal characteristics of UAV clusters, predict their future trajectories and identify potential intrusion intentions, providing early warning support for air defence systems. This study provides a new technical idea for UAV cluster intent recognition and promotes the research process of intelligent development of air defence system.