The Information Processing In Wireless Sensor Networks

Wireless sensors with the ability of sensing, computation and communication can work in the ad hoc mode to compose a wireless sensor network (WSN). WSN can monitor the targets and collect the information of the environment through the cooperation of large amount of sensors. This thesis concentrates on the information processing in wireless sensor networks and especially focuses on three aspects, i.e. capacity analysis, information-driven sensor selection algorithm and distributed source coding.Capacity analysis aims to explore the relation between the network capacity and the number of sensors, as well as that between the network capacity and the energy consumption, etc. under the constraint of the limited power per sensor. A scheme is proposed to improve the network capacity of densely deployed WSN by utilizing the low cost communication between neighboring sensors and the correlation of their data. The source firstly transfers its data to the neighboring sensors, which then forward the received data to the sink node, acting as the virtual antennas. Several sinks are deployed to alleviate the bottleneck effect of a single sink and the spatial-reuse of their frequencies can improve the capacity of such multi-modal network. For multi-hop networks, the relation between the number of sensors and the link capacity is analyzed, and the same capacity scaling law as that with the virtual multi-antenna technique is achieved by scheduling the transmission order of different nodes in the MAC layer. To meet with the energy limitation requirement of sensors, a scheme of optimizing the deployment of sensors and adopting the strategy of routing while compressing is proposed to minimize the network energy consumption.In our information-driven sensor selection algorithms, the information utility function is adopted to evaluate the information contribution of sensors. For the application of routing the user's query from the query proxy to the designated exit node, the sensors on the routing path perform sequential Bayesian filtering with the received information content and its own measurement to estimate the target status under the constraint of communication cost limitations. And then they select the neighboring sensor with the maximum information contribution as the next hop in the ellipse with the focuses of the current node and the exit node. The update of the target state and the sensor selection process repeats until arriving at the exit node.In the target tracking application with routing holes, an improved M-hop-neighbor-searching routing method is proposed, i.e. the sensor on the route evaluates the expected mutual information of all its neighboring nodes within M hops...