Compressive data aggregation in wireless sensor networks using sub-Gaussian random matrices

In this paper, we study a data aggregation problem in wireless sensor networks. We propose a Compressive Sensing (CS) based strategy which is able to reduce energy consumption and data collection latency. We adopt a random sensing matrix with entries drawn i.i.d. according to strictly sub-Gaussian distributions. Such a matrix have property such that a fraction of its entries are equal to zero with high probability. This enables us to collect data from only a fraction of the network without affecting data recovery, which helps reduce communication overheads. Linear networks and planar networks are considered. We compare the energy consumption and latency performance of our strategy with those of Compressive Data Gathering (CDG) scheme. Analytical and simulation results show that our scheme can reduce up to 44% and 67% of the energy consumption for linear and planar networks respectively, when the number of nodes is large. A significant improvement on the latency performance is observed as well.