BFVP: A Probabilistic UHF RFID Tag Localization Algorithm Using Bayesian Filter and a Variable Power RFID Model

Abstract

We present Bayesian filter of variable RF transmission power (BFVP), a Bayesian filter-based algorithm and a variable power radio-frequency identification (RFID) model for localizing passive ultra high frequency (UHF) RFID tags in complex environments, such as distribution centers/warehouses and retail sales floors. It enables a commercial off-the-shelf (COTS) RFID-equipped robot to provide the precise locations of passive UHF RFID tagged items. First, the robot navigates and fully covers a target space to collect observation of RFID tags using a COTS reader at variable RF transmission power. Every observation is associated with a pose, where the robot receives the response, and the RF transmission power. When the robot collects all responses from an RFID tag, BFVP can estimate the location of the tag, hence, localizing the tagged item. We tested the performance of BFVP in a mock apparel store; it exhibits less than 0.5-m localization error in the practical retail environment with significant multipath fading. Our proposed BFVP could enable promising applications that could greatly improve the efficiency of supply chain management by providing the precise locations of RFID tagged items.