Research On Sensor Placement And Energy Conservation For Wireless Sensor Networks

Recent advances in integrated sensing, micro-processing and wirelesscommunication have made Wireless Sensor Networks(WSNs), wirelessly self-organizedby a vast number of sensors, a reality in a variety of applications ranging from battle-field surveillance to environment observation and habitat monitoring and so on. TheWSNs used in the context of environmental monitoring, in conjunction with WWW,lead to the emergence of Sensor Web, which can provide human with one real-timeaccess interface to the sensed objects.Soil moisture is one of the most important monitoring objects for Earth Science.Therefore, we particularly studied the sensor placement of the Sensor Web based in-situ soil moisture sensing. As sensors are energy-restricted, they usually work in energy-efficient and non-continuous mode, irrespective of monitoring applications. However,such an energy-efficient working mode definitely leads to somewhat negative impactson the networks' monitoring quality such as data fidelity, coverage degree, packet delayto name a few. Thus, we investigated how to design and analyze different energy-savingpolicies satisfying with different monitoring requirements (e.g. more precise field data,higher coverage degree, lower packet delay etc.). In brief, our primary contributions aresummarized as follows:(1) We mainly studied the optimal sensor placement and field estimation in thecontext of in-situ soil moisture sensing. Combing the variation characteristics ofspatial soil moisture and the assumption of spatial Gaussian Process, we explored onescalable and robust cluster-based sensor placement algorithm behaving empiricallyoptimal performance, and specifically denoted that the cluster-based maximum mutualinformation criterion is a constant factor approximation of the optimal solution.(2) We primarily discussed the measurement scheduling and estimation usingcompressive sensing for in-situ soil moisture sensing. In order to obtain the underlyingtemporal evolution of soil moisture, we proposed one energy-efficient open-loopmethodology based on the sparse measurement scheduling and exact recovery algorithmusing compressive sensing theory. On the contrary, we raised a close-loop method thorough formulating the original problem as a partially observable Markov decisionproblem as well.(3)We systematically investigated the relationship between the k-coverage and thedegree of energy-saving for the traditional dense sensor networks. Building theprobability model for the k-coverage networks, we theoretically induced the energysaving in respect with coverage and node density in one2-coverage sensor networks.(4) We alternatively studied the problem of minimizing the packet delay in thedelay-tolerant sensor networks for one energy-conservation level. Rather than solvingthe original minimum problem, we constructed its corresponding dualversion aiming at maximizing one objective function, and then proposed one near-optimal greedy algorithm to this new problem.