Full-Dimension Relative Positioning for RFID-Enabled Self-Checkout Services

Self-checkout services in today's retail stores are well received as they set free the labor force of cashiers and shorten conventional checkout lines. However, existing self-checkout options either require customers to scan items one by one, which is troublesome and inefficient, or rely on deployments of massive sensors and cameras together with complex tracking algorithms. On the other hand, RFID-based item-level tagging in retail offers an extraordinary opportunity to enhance current checkout experiences. In this work, we propose Taggo, a lightweight and efficient self-checkout schema utilizing well-deployed RFIDs. Taggo attaches a few anchor tags on the four upper edges of each shopping cart, so as to figure out which cart each item belongs to, through relative positioning among the tagged items and anchor tags without knowing their absolute positions. Specifically, a full-dimension ordering technique is devised to accurately determine the order of tags in each dimension, as well as to address the negative impacts from imperfect measurements in indoor surroundings. Besides, we design a holistic classifying solution based on probabilistic modeling to map each item to the correct cart that carries it. We have implemented Taggo with commercial RFID devices and evaluated it extensively in our lab environment. On average, Taggo achieves 90% ordering accuracy in real-time, eventually producing 95% classifying accuracy.