| Abstract: | In a sensor network with a large numberof densely populated sensor nodes, a single target ofinterest may be detected by multiple sensor nodessimultaneously. Data collected from the sensor nodes areusually highly correlated, and hence energy saving usingin-network data fusion becomes possible. A traditionaldata fusion scheme starts with dividing the network intoclusters, followed by electing a sensor node as clusterhead in each cluster. A cluster head is responsible forcollecting data from all its cluster members, performingdata fusion on these data and transmitting the fused datato the base station. Assuming that a sensor node is onlycapable of handling a single node-to-node transmissionat a time and each transmission takes T time-slots, acluster head with n cluster members will take at least nTtime-slots to collect data from all its cluster members. Inthis paper, a tree-based network structure and itsformation algorithms are proposed. Simulation resultsshow that the proposed network structure can greatlyreduce the delay in data collection. |
