The vision of Internet of Things (IoT) is to embrace many enabling technologies for massive connectivity, adequate communication and powerful sensing capabilities for most items in the environment. When Kevin Ashton first coined this term, IoT was expected to collect and process massive amounts of data into an overarching network. The interplay between the Internet and the physical world will be made possible by the marriage of advanced communication and sensing technologies, which generate useful information to make important decisions and improve the quality of life. Among the many carriers of sensing technologies, radio signals have been extensively studied due to the unique advantages in propagating through obstacles, requiring no lighting conditions and protecting privacy. However, radio signals encounter many challenges that may not be present for other technologies, such as channel fading due to multi-path interferences, ambient dynamics, near-field coupling, and sparse spectrum resources.This dissertation contains two major research areas for IoT applications. First, the problem of indoor detection, locating and counting of device-free occupants by ambient radio markers will be discussed, which is based on the radio signals propagating through the multi-path environment differentiated by the occupants. Physics-based forward problems were formulated, and inverse solutions were developed to generate 3D voxel images of the indoor capture volume, from which the occupant locations and postures are determined. A machine learning model was also developed to classify the number of indoor occupants. The systems were prototyped by commercial off-the-shelf (COTS) radio-frequency identification (RFID). Second, the research direction of precision 3D localization of passive radio markers, such as RFID tags, will be discussed in detail. For structural integrity applications, these markers are often inside building materials. To couple sufficient radio energy to the marker inside a media with high and inhomogeneous permittivity, near-fields and multi-path effects have to be considered. Additionally considering the limited bandwidth resources and wavelength ambiguities due to phase rotations, we present a novel localization framework with curve fitting to accommodate nonlinear and non-monotonic phase-distance relations and ambiguity-free localization algorithms leveraged by channel redundancy from spatially diverse sensing units. Abundant channel resources are available from multiple-input multiple-output (MIMO) network, with aligned phase measurements that generate repeatable results over multiple device startups. The experimental system was prototyped by MIMO Universal Software Radio Peripheral (USRP) platform and harmonic RFID. For both research directions, the studies in this dissertation are expected to provide new insights and lay foundations for future research endeavors in the field of RF sensing and communications.

物联网 (IoT) 的愿景由许多技术赋能，为环境中的大多数物品提供大规模连接、足够的通信和强大的传感功能。当凯文·阿什顿（Kevin Ashton）首次创造这个术语时，人们期望物联网能够收集和处理大量数据并将其置于一个总体网络中。先进通信和传感技术的结合将使得互联网与物理世界之间的相互作用成为可能，而这些技术产生有用的信息来帮助人们做出重要决策并提高生活质量。在众多传感技术载体中，无线电信号由于具有穿越障碍物、无需照明条件、保护隐私等独特优势而受到广泛研究。然而，无线电信号遇到了其他技术可能不会出现的许多挑战，例如多径干扰、环境动态导致的信道衰落、近场耦合和稀疏频谱资源。本文包含物联网应用的两个主要研究领域。首先，文章将讨论使用环境无线电标识对无设备目标进行室内检测、定位和计数的问题。该问题建立在从多路径环境传播的无线电信号中区别出目标的反射信号。我们构建了基于物理规则的正向问题，并开发出逆问题解法来生成室内的三维体素图像，从而确定目标的位置和姿态。我们还开发出一个机器学习模型来对室内目标的数目进行分类。该系统采用商用现成 (COTS) 射频识别 (RFID) 作为原型验证。 第二，我们将详细讨论无源无线电标识（例如RFID标签）的精确三维定位的研究方向。对于结构完整性检测的应用场景，这些标识通常位于建筑材料内部。为了将足够的无线电能量耦合到位于介电常数高且不均匀的介质内的标识，我们必须考虑近场和多径效应。此外，考虑到有限的带宽资源和由于相位周期性旋转引起的整数波长的模糊性，我们提出了一种新型的定位算法框架。该框架具有曲线拟合功能，以适应非线性和非单调的相位-距离关系，以及基于空间多样分布的传感单元带来的冗余信道资源的无模糊定位算法。多输入多输出 (MIMO) 网络提供丰富的信道资源，并具有同步的相位测量结果，可在多设备多信道每次启动时产生可重复的结果。该实验系统采用 MIMO 软件无线电外设 (USRP) 平台和二次谐波 RFID 进行原型设计。 对于这两个研究方向，本文的研究有望为射频传感和通信领域的未来研究工作提供新的见解并奠定基础。