A propagation-centric transmitter localization method for deriving the spatial structure of opportunistic wireless networks

The design of emerging multi-tier dense wireless networks, which integrate opportunistically deployed devices into legacy infrastructure networks, will hugely benefit from a thorough understanding of the spatial structure of existing large-scale unplanned deployments. However, detailed and precise datasets which would enable acquiring such knowledge are unavailable and non-trivial to generate. In this paper we develop a new localization method that focuses on the derivation of transmitter distributions over large areas, such as those of opportunistic Wi-Fi networks. While prior work has emphasized single-node localization through geometrical inference on points of visibility and involved only rudimentary radio propagation modelling, we combine these two approaches in a hybrid technique to improve localization accuracy. Through exploitation of location information for a subset of known transmitters, we derive the parameters of a statistical propagation model that reduces the error induced by shadowing in realistic environments. After deriving theoretical bounds on propagation estimation from measurements, we compare our method to centroid-based localization approaches via simulation. We further assess our algorithm using data from a real Wi-Fi measurement campaign we have carried out in a suburban environment. Our results indicate a significant improvement in localization accuracy using our hybrid approach compared to conventional geometry-focused techniques.