Localization and Tomographic Imaging for Spatially Aware Mobile Radio Networks
Beck, Brian Michael
MetadataShow full item record
Mobile wireless sensors and networks have many potential applications, and have attracted much industry and academic interest in the past decade. This work pursues research which enhances the spatial awareness of such networks. Specifically, in this work we focus on original contributions to the spatial topics of interference/spectrum mapping, cooperative ultra-wideband (UWB) localization and tracking, uncooperative emitter localization, and active mapping of the radio frequency (RF) shadowing environment. Toward contributions in these areas, we introduce the Cognitive Spectrum Operations Testbed (CSOT), which was designed specifically for spatially oriented wireless network research. The system consists of eight mobile nodes, each with software defined radio (SDR) and UWB capabilities. The CSOT’s hardware and software are first demonstrated in an interference mapping experiment, which is used to optimize the positioning of a relay node. We then utilize CSOT to pursue high accuracy, cooperative indoor localization and tracking. This research utilizes fusion of both UWB ranging and odometry data. The experimental results demonstrate average positioning error of < 2 cm, and drops the requirement of pre-surveying the positions of anchor nodes. With precise localization in hand, we then use CSOT to explore the field of radio tomographic imaging (RTI). For our RTI work, UWB signal strength data is employed to produce accurate images of the RF shadowing environment. A major contribution of this work is its deployment in a completely uncalibrated network. All necessary parameters are estimated directly from the data. We show performance improvements through simulation and experiment, and via comparison with other techniques. Finally, we apply the uncalibrated network principle to the uncooperative emitter localization problem. We develop algorithms to model the effects of uncalibrated sensors on the data covariance and mitigate them. Further simulations and experimental results demonstrate significant localization performance improvements over naive methods.