Accurate real-time UAV flight-mode classification
Nikolaos Georgiou, Panayiotis Kolios · 2022
Identifying flight patterns in unmanned aerial vehicle (UAV) operations is a critical function, especially when UAVs are used in safety-critical missions. UAVs are used in various applications and their increased penetration in the market over the recent past has motivated investigations of their safety and security. For instance, due to their highly-non-linear dynamics it is inherently quite difficult to monitor and quickly and reliably identify a change in their behaviour or a fault in their sensors. Clearly, not being able to detect changes in the operations could cause a failure or event the complete loss of a UAV.To avoid such events, in this work we propose a mechanism for detecting the operational flight modes of a flying UAV based on onboard sensor data. We study and analyze two complementary approaches to identify these flight modes and evaluated their performance on real data gathered from various flights performed by a multirotor UAV. We concentrated our study on developing models that are both fast and accurate so that they could be executed at run-time and produce reliable results.