Restriction lift date: 2027-12-31
AI-enabled chipless RFID sensing system for reliable IoT applications
Loading...
Files
Date
2024
Authors
Rather, Nadeem
Journal Title
Journal ISSN
Volume Title
Publisher
University College Cork
Published Version
Abstract
The Internet of Things (IoT) is growing rapidly, driving the need for innovative and sustainable solutions for wireless identification and environmental monitoring. Passive Radio Frequency Identification technology (RFID) has been a key wireless communication technology enabling IoT. Recent advances have paved the way for battery-less, chipless RFID (CRFID), which eliminates the need for an integrated circuit (IC) component on the tag. This PhD thesis presents a new design strategy for developing concentric rings-based polarization-insensitive CRFID sensing tags. The proposed tag design approach of exponential spacing results in an 88.2% higher tag data encoding capacity than conventional designs which incorporate uniform spacing of the resonant rings. This is coupled with the idea of using the innermost ring for capacitive sensing. The concept of using RCS nulls for data encoding is implemented to enable convenient and accurate sensing by the innermost ring. This is made possible by adding an extra ring at the tag’s outermost edge. To enable robust detection of these tags, Artificial Intelligence (AI) is integrated on the reader side, employing both machine learning (ML) and deep learning (DL) techniques for decoding RCS EM signatures. In this research, ML and DL regression modelling techniques are applied to a dataset of measured RCS data derived from large-scale automated measurements of custom-designed, 4-bit CRFID sensor tags. The robotic measurement system is implemented using the first-of-its-kind automated data acquisition method using an industry-standard robot. The results show that all the ML/DL models were able to generalize well, that is, the ability of a model to perform accurately on new, previously unseen data. However, the 1D-CNN DL models outperformed the conventional ML models in the detection of ID and sensing values. In another contribution, a 3-bit depolarizing CRFID tag is developed and enabled for surface and shape robust detection using AI. For the first time reported, the system was trained on a dataset of 12,600 EM signatures, capturing varying surface permittivity, tilt angles, read ranges, and tag bend scenarios. The AI models using 1D-CNN are trained and validated, resulting in a low RMSE of 0.040 (0.66%) for tag ID detection. On the same dataset, for the first time, DL models were evaluated with Bidirectional Long Short-Term Memory (Bi-LSTM) and attention mechanism, further reducing the RMSE to 0.029 (0.48%). The outcomes of this thesis contribute significantly towards state-of-the-art of AI-enabled CRFID systems for robust and reliable real-world IoT applications.
Description
Keywords
Chipless radio frequency identification (CRFID) , Convolutional neural networks (CNN) , Deep learning (DL) , Electromagnetics (EMs) , Machine learning (ML) , Radar cross section (RCS) , Radio frequency identification (RFID) , Robots
Citation
Rather, N. 2024. AI-enabled chipless RFID sensing system for reliable IoT applications. PhD Thesis, University College Cork.