Trajectory Prediction of UAV Swarms Based on Cooperative Graph Attention Networks

Accurate trajectory prediction can enhance the operational efficiency of UAV swarms, ensure flight safety, optimize resource allocation, improve the swarm's response capabilities, and increase the autonomy of the UAV swarm. This study models the UAV swarm operation system based on the Olfati-Saber flocking dynamics. Through dynamic interactions within the system, an adaptive cooperative graph for the UAV swarm is established, integrating the prediction results from multiple cooperative nodes. On this basis, the mutual influences within the swarm are considered to make accurate predictions for individual UAV trajectories. Additionally, an online learning mechanism is introduced, enabling the model to adapt in real-time to changes in swarm states. Experimental results indicate that this model can effectively predict the flight trajectories of UAV swarms, significantly improving prediction accuracy compared to traditional prediction models.