RSS-Based Localization Techniques with Large-Scale Experimental Evaluation

Signal strength-based cooperative localization has gained significant interest due to its low complexity and cost-effectiveness. The conventional techniques often use Taylor expansion to address the non-convex, non-linear, and discontinuous characteristics of the maximum likelihood (ML) objective function. However, this results in a considerable residual error that exponentially grows with noise. To address this issue, we propose a cooperative (and non-cooperative) localization technique using received signal strength (RSS) measurements, named C-UA (and NC-UA), which employs a relative error-based estimator. We then use weighted non-linear least squares (NLS) to formulate a semidefinite programming (SDP) problem. C-UA and NC-CA can jointly estimate the location and transmit power of target nodes while considering the uncertainty in anchor nodes' locations. We also derive the Cramer-Rao lower bound (CRLB) involving unknown transmit power and anchor location uncertainty. We perform extensive outdoor real-world tests in an open field (640 m × 180 m), using 50 Bluetooth transceivers to collect RSS measurements. We record the location of each node using two devices: a real-time kinematic (RTK)-GPS system, providing centimeter-level accuracy, and a standard GPS device, offering meter-level accuracy. The RTK-GPS measurements are considered the true positions, while the standard GPS measurements, with lower precision, are used to represent the uncertainty in the locations. Through extensive simulations and experimentation, we demonstrate the superior performance of C-UA and NC-CA over existing localization techniques.