UC Santa Cruz

With the advances of new technologies such as Radio-frequency identification (RFID) systems, pervasive sensing, and cyber-physical systems (CPS), the Internet-of-Things enables inter-connection of small objects (such as tagged items, embedded systems and mobile devices) and data collection, delivery, as well as processing among them. This thesis focuses on integrating large-scale and low-cost passive RFID tags into existing pervasive-sensing applications to enable multi-functional and cost-efficient IoT systems. We believe the inherent advantages of RFID systems to identify, trace, and track information using easily deployable tags provide unique opportunities to enable many novel IoT applications in new areas of sensing, actuation, and user interaction, which is far beyond its traditional use in supply chain management. However, the design of these applications involves challenges, as the limited computation ability and simple functionality of passive tags may make them ill-fitted for meeting the diverse requirements. To tackle these issues, this thesis takes a deep look at exploring the potential of commercial off-the-shelf (COTS) passive tags, and proposes new applications in the following themes: • Enabling cyber-physical connection using RFID passive tags, where system can seamlessly detect user-item interaction and gather information from real-world subjects. • Fusing the information from computer vision and RFID sensing channel to actively find/track the mobile object/person with least training efforts. • Addressing scalability issue and domain shift challenge in human gesture recognition with RFID via employing domain-adaptive few-shot learning (DA-FSL).