3D Localization of RFID Antenna Tags Using Convolutional Neural Networks

With the Internet of Things (IoT) becoming widespread, there has been a rising demand for indoor location-based services. In recent trends, radio frequency identification (RFID) has become an integral part of the production of IoT. Conventional methods use prior knowledge of antenna and tag positioning along with high-precision equipment capable of collecting phase or time-of-arrival data for robust estimation of 3-D location. In this work, we propose a 3-D localization method based on deep learning that relies on the phase and received signal strength indicator (RSSI) captured by steering beams to various locations using a phased array antenna. We evaluate the efficiency of this system by estimating the 3-D location of seven RFID tags mounted on metallic surfaces placed in a naturalistic environment. To evaluate the generalization of the proposed approach, we cross-validate the localization performance in different environments. The localization performance of the proposed approach is also tested on different form factors of the RFID tag. With no prior information of either the tags or environment, the proposed system was able to achieve an average localization error as low as 1.33 cm with better system stability.