Contents I. Introduction
Identifying and localizing objects is a primary application driver for the Internet of Things. Such capability can be utilized in many ways, such as for merchandise security, warehouse management, and smart spaces. For example, with regards to smart spaces, detecting the position of objects on a table enhances machine-learning algorithms for perception of human activity, since object interactions are strong features of the activity being performed [1], [2]. Currently, the primary technology for object localization is vision-based sensing. This is limited by line-of-sight, requires extensive training data, and is computationally expensive. A more direct approach is to distribute sensors over a large area, embedding them in everyday surfaces, to locally detect objects, wherever they are placed. While the range of technology tradeoffs impacting such applications are still being investigated, in [3] it is shown that such an approach can significantly improve the learning efficiency and recognition accuracy for human-activity perception. A potential technology for “sensing” an object’s identity is RFID, and the tight read range of 13.56 MHz HF RFID can provide effective localization. Fig. 1(a) illustrates an envisioned application, where a table lined with a thin, dense array of RFID readers unobtrusively detects tagged objects placed on its surface.
System concept and architectures. (a) Concept of object-detection/localization. (b) RFID reader array using direct address scheme. (c) Array with TFT-switch-based active matrix control. (d) Proposed active matrix design, with oscillations produced locally by RDR-PIXELs.
