Compressed RF tomography: Centralized and decentralized approaches .

摘要

Radio Frequency (RF) tomography refers to the process of inferring information about the contents of an environment via capturing and analyzing transmitted RF signals. Received Signal Strength (RSS) measurements acquired by the sensor nodes are analyzed to determine the location of certain obstructions in the environment. Therefore, a wireless sensor network can employ RF tomography for surveillance and monitoring in a environment. In this thesis, we introduce Compressed RF Tomography for monitoring via wireless sensor nodes, which requires fewer RSS measurements than non-compressed RF tomography, allowing for an extended network lifetime. Compressed sensing is a recent field of research that has captured considerable attention in engineering due to its efficiency in signal sampling. Combined with RF tomography, it introduces a new approach to monitoring in wireless sensor networks. Our main contributions in this work include employing compressive sensing techniques in RF tomographic imaging, and demonstrating their capabilities in centralized and decentralized processing schemes. We present an approach that uses battery power more efficiently and performs better when only few sensors are available. Moreover, we introduce decentralized schemes for in-network data analysis. This allows sensors to cooperatively monitor the environment without the need for a fusion center. Simulations throughout the thesis illustrate the performance of our approach under different situations. Real sensor network data is also used to compare our approaches with the existing approach.