To Örebro University

Precise gas concentration measurements are often difficult, especially by in-situ sensors mounted on an Unmanned Aerial Vehicle (UAV). Simple gas detection, on the other hand, is more robust and reliable, yet brings significantly less information for gas source localization. In this paper, we compensate for the lack of information by a physical model of gas propagation based on the advection-diffusion Partial Differential Equation (PDE). By linking binary gas detection measurements to computed gas concentration using the physical model and an appropriately designed likelihood function, it becomes possible to identify the most likely gas source distribution. The approach was validated in two experiments with ethanol and smoke as "toy" gasses. It is shown that the method is able to successfully localize the source locations in experiments based on gas detection measurements taken by a UAV.