Energy-balanced compressive data gathering in Wireless Sensor Networks

Compressive Sensing (CS) can use fewer samples to recover a great number of original data, which have a sparse representation in a proper basis. For energy-constrained Wireless Sensor Networks (WSNs), CS provides an effective data gathering approach. Gaussian random matrix satisfies Restricted Isometry Property (RIP) with high probability. The class of matrices is usually selected as the measurement matrix for compressive data gathering in WSNs. However, they are dense, and the computational complexity is higher. On the other side, sparse binary matrix with a fixed number of nonzero entries in each column satisfies RIP-1 property. Due to the higher sparsity, the class of sparse binary matrix is chosen as the measurement matrix in the paper. In order to adapt to the dynamic change of network topology, we design a mobile agent based compressive data gathering algorithm (MA-Greedy algorithm), where each sensor node is uniformly visited in M measurements. Coefficient of Variation (CV) is proposed to evaluate the balance of energy consumption. The numerical experiments show the proposed algorithm is superior to other algorithms (i.e. non-CS, plain-CS, Hybrid-CS, and Distributed Compressive Sparse Sampling (DCSS)) in terms of energy balance. Moreover, we discover the performance of reconstructing sparse zero-one signals by sparse binary matrix, which is used in the proposed MA-Greedy algorithm, is better than that by Gaussian random matrix when Basis Pursuit (BP) algorithm is used for signal recovery.