Phase-based Device-free Tracking exploiting a Cylindrical Human model and Kalman Smoothing

This work proposes a phase-based device-free lo-calization scheme (DFL) that tracks a single human target by exploiting a network of passive RFID tags inside the monitored area. As the target moves through the propagation environment, it obscures some antenna-to-tag links and influences the phase measurements of the backscattered field. By comparing them with measurements collected in a target-free environment, it is possible to identify the obscured tags and consequently, the antenna's shadow zone. Such information is then directly mapped to the target's location by employing a geometrical model, according to which the target holds a cylindrical volume. The shadow's boundary lines should be tangent to the cylindrical surface and a system of linear equations is employed to identify its centroid. Localization is also feasible even when a single antenna is available. A two-stage (forward and backward) Kalman filter is developed to constantly update the target's location over time. Experimental results employing either a single or multiple antennas report a median error of about 15cm.