Improved Package Orientation Estimation Through RFID Signal Analysis with LightGBM

Radio-frequency identification (RFID) technology is widely used across industries due to its affordability, low power requirements, and compact form factor. Its adaptability has led to expanded research into applications beyond traditional tracking. This study investigates the use of commercial RFID antennas to determine package orientation by analyzing Phase and RSSI readings. Using machine learning models, we evaluated sixteen unique configurations and package orientations. Results show that the optimal placement of an RFID tag can achieve up to 92.3 % accuracy in orientation estimation, with the most effective setup utilizing a single antenna and tag. This research presents an innovative approach to posture detection, underscoring RFID's potential for advanced tracking and monitoring applications.