Following an RF trail to its source

We consider the problem of drawing a mobile platform such as an Unmanned Aerial Vehicle (UAV) to a radio frequency (RF) source surrounded by local scatterers. The UAV starts from a significant distance (e.g., kilometers) from the source, and uses received signal strength (RSS) measurements along its path to continue to adapt its trajectory. The fundamental bottleneck in using the RSS for this purpose is that large spatial variations in the the RSS due to multipath fading can swamp out the relatively smaller variations due to the change in the distance to the source. We characterize these variations using ray tracing, explicitly modeling the scattering environment around the source. We discover that the rate of spatial variations decreases as the direction of travel becomes more aligned with the line-of-sight (LoS) to the source. Thus, it is possible to get trapped in a very long fade which makes it difficult to gather information regarding distance to the source. We use these insights to devise a multi-phase algorithm which first uses averages of RSS in multiple directions to estimate the "right" direction of travel towards the source, and then uses longer term averages to decide how long to proceed along the chosen direction. The distance traveled by the UAV to get to the source is, on average, about three times longer than the shortest path. While the proposed algorithm is inspired by the motion of bacteria in search of food, it has been significantly optimized to take into account the sensitivity of RSS to direction of travel.