An Intelligent Framework for Prediction of a UAV’s Flight Time

The success of an unmanned aerial vehicle's (UAVs) or drone's mission is contingent upon planning, command, control, tasking, and communications. Drone-mounted payloads impact UAV flight time. Depending upon the payload, flight times may vary. As such, it is important to know beforehand the expected flight time of a drone, in order to ensure a successful flight. Currently, there are no methods or algorithms for calculating or predicting a drone's flight time that takes into account varying payloads. Manufacturers' technical manuals usually provide only the best-case predictions of UAV flight time; these assume no additional mounted items (e.g., onboard wireless sensor modules, cameras, etc.). In this paper, we describe an empirical study of UAV flight events, and propose regression and deep learning (DL)-based methods to predict accurately the flight time of UAVs. Our methods take onto account both the payload weight and energy dissipation from the onboard battery. The payloads used in our study include gimbal-mounted RGB/thermal cameras, as well as onboard computers. It is expected that this work will provide important guidance to researchers and planners of UAVs.