Multirotor unmanned aerial vehicles (UAVs), or drones, are increasingly being used to spray liquid pesticides to control emerging pest infestations in field crops. In recent years, UAVs have been used to release predatory mites and other natural enemies to optimise and promote sustainable pest management practices by relying less on conventional insecticides. Drone dispensed samples of predatory mites are typically mixed with a granular material, vermiculite, which serves as a filler. The low density of the vermiculite and weather conditions (mainly wind), influences the distribution pattern of predatory mites when delivered by a UAV-based system. The purpose of this paper is to present a data-driven methodology to develop a mathematical model that can be used to optimise UAV-based autonomous dispensing of predatory mites. The model characterises the distribution of vermiculite as a function of wind speed and direction, and the UAVs altitude and forward speed. The model is constructed by first conducting outdoor experiments and then using machine-learning techniques on the collected data. The constructed model produced an average generalisation error of 12.8%, RMSE. Due to its parametric and predictive nature, the model is amenable for the future design of UAV flight controllers that can compensate for the targeting error caused by wind. The proposed modelling methodology could be useful not only for the dispensing of predatory mites, but also for other UAV dispensing applications, such as liquid or granular pesticide deliveries.Keywords: unmanned aerial vehicle, precision pest management, machine learning, natural enemies, precision agriculture, predatory mites.
NomenclatureUAV unmanned aerial vehicle X the direction perpendicular to UAV flight path RMSE root mean squared error wx wind speed in the X direction, km h -1 ℎ UAV altitude, m wy wind speed along the flight path, km h -1 x lateral offset in the X direction, m R 2 statistic, coefficient of determination vermiculite density, g m -2 wind speed, km h -1 UAV forward speed, km h -1 wind direction with respect to UAV heading, °