Digital Twin of Retail Environment Using RFID Particle Filter Localization

Abstract

Localization using RFID technology enables a realistic and cost effective solution for improving inventory accuracy in varying industries. The retail environment being an ever growing topic of interest for RFID implementation, a proposed method of Bayesian statistical algorithms centered on particle filter localization is analyzed. With the use of stereo tracking cameras, a 3D point cloud of the environment is created in real time to enable localization of various items. This method is built upon a previous work that does not rely on RSSI or phase measurements, instead the proposed method leverages a fixed RF transmission power model enabling localization capabilities to all existing RFID devices. After the environment is transferred to a Digital Twin using RGB-D cameras, the position of various RFID tags are estimated. To determine the performance of the proposed particle filter, a commercial off the-shelf (COTS) storage rack is used to emulate a retail environment. A testing procedure is conducted to compare the proposed particle filter approach with the previous Bayesian filter. The proposed scheme performs adequately in heavy-clutter environments with a reduced computational time.