Multiple transmitter localization using clustering by likelihood of transmitter proximity

We propose a measurement clustering method for estimating the locations of multiple radio transmitters based only on received signal strength and a performance analysis approach drawn from a hypothesis testing framework. Measurements are collected at a set of arbitrary locations in the area of interest. Measurements are retained (or discarded) for transmitter localization based on the probability that a transmitter lies within a given neighborhood of the measurement location. Retained measurements are clustered using k-means or a similar approach, and individual transmitter localization is performed in each cluster. The proposed approach makes no assumptions that the measurements are clustered around transmitters; measurement locations may be at a similar distance from several transmitters, complicating the estimation process. The effect of the choice of neighborhood size and measurement threshold is analyzed using a hypothesis testing framework in which the probability of detection, probability of false alarm, and receiver operating characteristic are derived. Simulation results show that, over a wide range of shadowing noise variance, the proposed clustering-based localization achieves performance gains over algorithms that estimate transmitter locations jointly.