AWG-Detector: A machine learning tool for the accurate detection of Anomalies due to Wind Gusts (AWG) in the adaptive Altitude control unit of an Aerosonde unmanned Aerial Vehicle

Use of unmanned Aerial Vehicles (UAVs) has gained significant importance in the recent years because of their ability to remotely monitor and perform various tasks in an autonomous manner. However, the control unit of such UAVs fails to adapt quickly when the UAVs are exposed to unpredictable and violent external disturbances such as violent wind gusts and extreme weather conditions. The cost of such adaptation failures can be extremely high and therefore, in order to use any crash preventing strategy, it is imperative to design and use intelligent tools for the early detection of such failures. In this paper we present a machine learning based autonomous tool - AWG-Detector - that detects Anomalies due to Wind Gusts (AWG), in our adaptive Altitude control unit of an Aerosonde UAV. This adaptive Altitude control unit comprises of a PI based Roll controller and a Hybrid neuro-fuzzy based Pitch controller. Experimental results show that our AWG-Detector achieves an accuracy of more than 99% in detecting anomalies due to wind gusts. To the best of our knowledge, this is the first study that targets the detection of Wind Gust anomalies in the Altitude control unit of an Aerosonde UAV by developing a comparison of five well-known machine learning techniques.