Deterministic binary matrix based compressive data aggregation in big data WSNs

In big data wireless sensor networks, the volume of data sharply increases at an unprecedented rate and the dense deployment of sensor nodes will lead to high spatial-temporal correlation and redundancy of sensors' readings. Compressive data aggregation may be an indispensable way to eliminate the redundancy. However, the existing compressive data aggregation requires a large number of sensor nodes to take part in each measurement, which may cause heavy load in data transmission. To solve this problem, in this paper, we propose a new compressive data aggregation scheme based on compressive sensing. We apply the deterministic binary matrix based on low density parity check codes as measurement matrix. Each row of the measurement matrix represents a projection process. Owing to the sparsity characteristics of the matrix, only the nodes whose corresponding elements in the matrix are non-zero take part in each projection. Each projection can form an aggregation tree with minimum energy consumption. After all the measurements are collected, the sink node can recover original readings precisely. Simulation results show that our algorithm can efficiently reduce the number of the transmitted packets and the energy consumption of the whole network while reconstructing the original readings accurately.