Accurate Device-Free Localization with Little Human Cost

Abstract

The device-free localization (DFL), i.e., localizing target without requiring target attached any devices, is attractive. Current localization methods, however, query a significant degree of pre-deployment effort, such as the transceivers' locations, the transmission power, which cost huge human effort. In this paper, we present Alico, an accurate and low human cost DFL method that does not require any pre-deployment effort, such as building the detailed fingerprints or requiring the prior knowledge of deployment. The key intuition is that (i) the distorted wireless links caused by the target, even the many from unknown locations, are constrained the presence of the target; (ii) with the increase of the number of unknown targets and transceivers, the constraints grows in a quadratic fashion, while the unknown locations of targets and transceivers grows linearly. This suggests that given enough measurements, there will be eventually enough constraints to make the every target uniquely localizable. Alico leverages these constraints and model them as a set of equations. By using a hybrid gradient descent and genetic algorithms, Alico can solve the equations and estimate the target locations accurately based just on the Received Signal Strength (RSS) measurements. Despite the absence of any explicit pre-deployment calibration effort, Alico achieves the 60th and 80th percentile errors of 1m and 1.4m in real-world experiments, respectively, which is better than the three state-of-the-art algorithms.