SuPRANO: On the localization of connected vehicles in hostile environments: A Semi-suPervised manifold leaRning technique

Abstract

In wireless hostile environments such as tunnels, tall buildings, undergrounds and dense vegetation where Global Positioning System (GPS) signals can be unavailable, vehicles are prevented from exchanging accurate positions. Hence critical information may be lost or misled. To overcome these limitations, this paper proposes an innovative technique for localization estimation called SuPRANO, a Semi-suPervised manifold leaRning based locAlization algorithm for vehicular NetwOrks. The key innovation in our technique is leverage the theory of semi-supervised learning. Specifically, SuPRANO employs a certain number of well localized vehicles, called leading vehicles, that collect signal measurements from non-localized vehicles (non leading vehicles) to estimate the position of these latter. The resulting technique is naturally realistic and performs very well.