Semi-supervised Localisation utilizing CSI at Large Antenna Array Base Stations

In this paper we present a localization method that exploits intrinsic properties of channel state information (CSI) acquired at a massive MIMO base station. This method combines dimensionality reduction techniques to extract spatial information with trilateration on K-reference locations to estimate the position of the transmitter. We show that the map constructed by dimensionality reduction techniques can be utilized to achieve different levels of accuracy solely from CSI samples from different transmitters depending on the density of reference point locations.