Artificial Intelligence enhances Smart RFID Portal for retail

Abstract

This paper investigates a low-cost implementation of a UHF-RFID portal enhanced by artificial intelligence to assess whether a tag is crossing the portal or it is static in the surrounding. The reference scenario concerns anti-theft systems in fashion stores, where the use of shielded RFID portals is not feasible and spurious tag readings must be filtered by alternative methods. The portal consists of a commercial RFID reader connected to an array antenna placed at the store entrance to monitor the tag crossing. Data processing involves Received Signal Strength Indicator sequences and a comparison among Support Vector Machine (SVM) and several LSTM (Long Short-Term Memory) neural networks investigated to perform classification. System and algorithm validation are conducted through an experimental analysis.