SBIR Phase I: A Handheld Fine-Grained Radio Frequency IDentification (RFID) Localization System for Retail Automation

Cartesian Systems, Inc., Cambridge MA

Investigators

Abstract

The broader/commercial impacts of this Small Business Innovation Research (SBIR) Phase I project will protect retail stores from loss of merchandise. Brick-and-mortar retail is undergoing an unprecedented transformation, having lost billions of dollars over the past decade due to labor shortages, competition from e-commerce giants, and changing expectations from the modern consumer. To address these issues, retailers have been adopting new digital technologies to gain visibility into their inventory, optimize store operations, and gain customer insights. A key technology that has been adopted by over 90% of US retailers, is Radio Frequency IDentification (RFID). RFID tags are cheap, wireless, and battery-less stickers (similar to barcodes) that have allowed retailers to achieve accurate store-wide inventory, resulting in a significant revenue increase for retailers. In contrast to existing (portable) RFID technology which can only determine whether RFID-tagged items are in the store (i.e., inventory), the proposed technology aims to precisely locate these items throughout the store. The technology leverages billions of off-the-shelf ultra-high frequency (UHF) RFID tags that are already attached to clothing, footwear, and apparel items. In contrast to existing mobile solutions which can only detect RFID-tagged items, the team's handheld device leverages sophisticated signal excitation and processing techniques to pin down each RFID’s exact position with decimeter-scale accuracy. This SBIR Phase 1 project will build a system capable of identifying and precisely locating RFID-tagged items and includes three main innovative components: (1) a portable, handheld wireless device for locating RFIDs, (2) a scalable cloud and edge computing platform to process and store the data, and (3) a mobile and web user interface for accessing the data and optimizing picking tasks for retail store associates. Realizing the end-to-end platform requires developing efficient sensor fusion algorithms and low-power, low-cost hardware for accurate, robust, and low-latency localization. This technology necessitates addressing challenges that arise from the computational, memory, bandwidth, and power constraints on the edge device. The platform also requires developing the split and cloud computing architecture to efficiently process data from multiple handheld devices in real-time as well as provide the generalizable application programming interfaces (APIs) to integrate this data pipeline with the retail customers. By the end of the Phase I period, the project will have piloted the fully-integrated system in a retail store to evaluate its real-world performance. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.