UAV-FD: a dataset for actuator fault detection in multirotor drones

Multirotor drones are equipped with propellers that may get damaged in flight in case of a collision with an obstacle or a rough landing. In view of safety-critical applications, such as flying over crowded areas or future passenger drones, being aware of a damaged actuator becomes essential to enhance system integrity. Therefore, in this paper we present a public dataset, namely UAV-FD, where real flight data from a multirotor under the effects of a chipped blade are collected. A conventional ArduPilot-based controller is employed, where the ArduPilot firmware is customized to increase the signal logging rate of selected variables, thus capturing information at higher frequencies. Moreover, the actual speed of each motor is measured and made available. Finally, we provide an illustrative fault detection strategy, based on MATLAB Diagnostic Feature Designer toolbox, to show how the dataset can be used and the blade chipping can be detected.