Empirical Fading Model and Bayesian Calibration for Multipath-Enhanced Device-Free Localization

Multipath-enhanced device-free localization (MDFL) systems determine presence and location of objects and users not necessarily equipped with localization devices. For localization, MDFL systems exploit user-induced changes in the power of all received signal components, including both line-of-sight and multipath components (MPCs). In this work, we therefore provide a statistical fading model that describes user-induced changes in received power specifically for MPCs. The model is derived and validated empirically using an extensive set of wideband and ultra-wideband measurement data. Since the localization performance of MDFL systems strongly depends on the information about the propagation paths within the wireless network, we further propose a Bayesian calibration approach that estimates the location of the reflection points of MPCs caused by single-bounce reflections. For MPCs caused by single-bounce reflections, the solution space of possible locations of reflection points is constrained to the delay ellipse, which allows the formulation of a computationally efficient one-dimensional estimation problem. Eventually, the problem is solved by sequential Bayesian estimation. The applicability of the proposed approach is demonstrated and evaluated using measurement data. Independent of the underlying measurement system, the Bayesian calibration approach is shown to robustly estimate the locations of the reflection points in different environments. Finally, the localization results of MDFL for an indoor scenario confirm the applicability of the Bayesian calibration approach.