Indoor Area Estimation System Using RSSI-Measuring Handheld Reader Utilizing Directional Reference RFID Tags and Machine Learning

• Published in: IEEE access 2024, Vol.12, p.157872-157887
• Main Authors: Hadi, Danang Kumara, Song, Zequn, Rahmadya, Budi, Kozume, Shogo, Sumiya, Tomonori, Sun, Ran, Takeda, Shigeki, Wang, Xiaoyan
• Format: Article
• Language: English
• Subjects: Accuracy; Algorithms; Antenna radiation patterns; Cold storage; Data acquisition; Global Positioning System; Handheld reader; indoor area; Indoor environment; Indoor environments; Inventory management; Localization; Localization method; Location awareness; Machine learning; Measuring instruments; Radio frequency identification; Radio waves; Received signal strength indicator; Reflector antennas; RFID tags; RSSI; Signal strength; Supply chains; Tags; Warehouses; Wave propagation
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2024.3486792

Achieving efficient warehouse operations for product management in an indoor environment has recently become a challenging issue. If a user can store a product in a vacant place and then roughly localize the product with a simple system, this localization system will enable efficient and flexible area usage in a warehouse. Complex radio wave propagation phenomena also make indoor localization more challenging. This paper introduces a radio frequency identification (RFID) system to localize products in indoor environments, including warehouses and cold storage. This approach uses distributed directional reference RFID tags in the areas as product location beacons, enabling received signal strength indicator (RSSI) measurements reflecting information on distances and directions for determining product coordinates. A user with a handheld reader stands in the close vicinity of the product. Then, the user reads the surrounding reference RFID tags for collecting the RSSI data. The use of machine learning (ML) addresses unstable user behaviors and unexpected acquired RSSI variations due to wireless propagation. Regression and classification algorithms in ML estimate product locations. The experimental demonstrations in actual indoor environments validate the proposed localization method. The experimental environments measure 24 \,m \times 12 \,m in a conference room and 9 \,m \times 12 \,m in a laboratory room. Experimental results show that this approach can provide localization accuracy of less than 2 meters, with wide application potential in inventory management and product tracking in various indoor environments, including factories, warehouses, and cold storage.