UAV Anomaly Detection Using Fleet Data Based on DBSCAN-MSET Method

With the rapid development of artificial intelligence, unmanned aerial vehicle (UAV) has been widely applied in multiple fields for their convenience, low cost, and powerful function. Compared with single-UAV data, the UAV fleet data has more rich features, which can provide greater data collection and rich data information support to achieve more accurate anomaly detection. However, the flight paths of different UAVs vary greatly, which brings many challenges to fleet-level data analysis. And it is difficult for UAV fleet data to get exception labels to accurately build an anomaly detection model. Based on the above problems, this paper presents an anomaly detection method based on multivariate state estimation (MSET) and density-based spatial clustering of applications with noise (DBSCAN) optimization (DBSCAN- MSET). Establishing a normal state representation model from historical UAV data can effectively detect whether the following flight is abnormal. The proposed method is validated against the actual UAV data, and the results show the validity of our method.