Physical Model-based Calibration for Device-Free Radio Localization and Motion Tracking

Device-Free Localization (DFL) systems employ a wireless network of radio-frequency (RF) nodes to detect the presence, and locate, or track, the position of targets (people or objects) in a monitored area, without requiring the targets to be instrumented or collaborative. DFL algorithms rely on an initial calibration procedure to avoid problems related to the electromagnetic (EM) effects induced by the environment. This calibration phase includes finger-printing measurement steps that require the presence of the targets in known locations, namely landmark points. These steps are typically time consuming and error-prone since each RF node needs to perform several noisy signal measurements for the assessment of body-induced alterations of the radio propagation. Moreover, modifications in the environment involve frequent recalibration runs to maintain a good level of localization accuracy. We present here a novel calibration scheme based on a physical model that predicts the effects of body-induced alterations of the EM field using a small number of the landmark points used to tune the model parameters. The framework is based on the scalar diffraction theory and it is able to predict the Received Signal Strength at the receivers for an assigned position and size of the targets and the geometry of the links.