Virtualized Controller for Computational RFID-based IoT Sensors in Industry 4.0 436 views

RFID technology is ubiquitous in business and industrial operations, access security, and identification methods. With the advent of Computational RFID (CRFID), microcontrollers integrated into RFID tags add computational capabilities that enable broader adoption in the IoT. However, the challenge CRFID encounters is optimizing power distribution and consumption across the billions of IoT devices expected in the coming years. This dissertation explores approaches to balance RF energy harvesting and minimize power consumption in RFID-based IoT devices, addressing the crucial issue of battery limitations in terms of lifetime and maintenance. We focus on enhancing sensor tags beyond the role of simple data collectors into intelligent systems that utilize RF for sensing, computing, and self-power. The approach proposed reaches back to the core aspects of RFID and preserves the simplicity of RFID tags, shifting computational tasks from the tag microcontroller onto the reader to optimize tag resources. We implement a virtualized controller for SPI-over-RF, enabling the transmission of wireless SPI control instructions from the reader to embedded sensors on tags via the RF chip. We explore the development of a reader's firmware capable of handling custom RF SPI commands, creating a framework for the research, design, and fabrication of CRFID prototypes. Our research demonstrated that VCRFID sensing devices eliminate the need for a microcontroller and its associated power requirements, achieving a 97% reduction in energy consumption compared to tags with MCUs. With the proposed RF harvester sensitivity of -31.4 dBm and a power conversion efficiency of 31.3%, longer-range operations would extend the VCRFID's reach beyond the initially reported distance. We completed and analyzed the integration of the VCRFID system from an edge-powered wireless sensor network to a cloud platform. The combination of VCRFID sensing technology and machine learning methods promises to advance the capabilities of RFID-based wireless IoT sensors for predictive maintenance applications in Industry 4.0.

PHD (Doctor of Philosophy)

RFID; Internet-of-Things; Energy Harvesting; Machine Learning; Wireless Sensing

English

All rights reserved (no additional license for public reuse)

2024-04-24