Anomalous Identification in UAVs Based on Receding-Horizon Clustering

This paper presents a novel approach to identifying anomalous individuals among UAVs utilizing receding-horizon clustering. The artificial potential field methods (APF) with a leader-follower mechanism are employed to generate multiple intelligently designed anomalous individuals. Within the predefined time window, the trajectory features of each UAV are extracted, which specifically include spatial positioning and trajectory detection indicators, dynamic speed and acceleration features, as well as motion path consistency. Then, the k-means method is used for clustering. Simulation results show that the receding-horizon clustering technique enables real-time identification of anomalous individuals among UAVs. In 1000 experiments, the identification accuracy exceeded 95%, and the detection time was less than 3 seconds.