Airborne Sensor Data-Based Unsupervised Recursive Identification for UAV Flight Phases
Benkuan Wang, Xiyuan Peng, Datong Liu
Abstract
With the widespread application of Unmanned Aerial Vehicle (UAV), more and more attention has been paid to the analysis of flight data, which can realize the condition monitoring and help improve operational safety. The UAV flight phase is dynamically switched, and flight phase identification is the vital basis for accurate flight data analysis. However, UAV flight data do not always contain a parameter that can be used for the direct division of flight phases, and it is difficult to give uniform thresholds for the flight phase identification. To realize the automatic identification of UAV flight phases based on airborne sensor data, an unsupervised Gaussian Mixture Model Recursive Clustering (GMMRC) method is proposed. GMMRC divides a complex flight phase identification task into simple subtasks which are corresponding to UAV operations. It can not only improve the performance of identification, but also enhance the interpretability of the results. In the recursive method, GMM is used as an unsupervised identification operator to obtain the desired identification results in each subtask. To further improve the identification effect, an interval detection method is used to eliminate false results. Experiments based on simulated flight data and real small fixed-wing UAV flight data demonstrate the potential of this method for accurate and robust flight phase identification.