Analysis of Machine Learning Algorithms for USRP-based Smart Chipless RFID Readers

The classification performance of machine learning (ML) on measurements from a low-cost chipless RFID reader is examined. We first justify the need of replacing the laboratory equipment used for chipless measurements with cheaper and less bulky devices such as software defined radios (SDRs). We take reflection measurements of four circular ring resonator (CRR) tags with different IDs employing a custom programmed USRP N210 SDR. We describe the step by step workflow followed to train five different ML models and correctly identify the tag IDs with over 93.3% accuracy. Future directions of research are also discussed.