Node clustering for data collection in wireless sensor networks using graph-transform and compressive sampling

In this paper, we address the problem of node clustering for compressed sensing (CS) based data collection in wireless sensor networks (WSNs). With consideration of recovery accuracy, communication cost and residual energy, two clustering strategies are proposed. Both strategies utilize Lapacian eigenvectors corresponding to the topology graph as a sparsifying basis, termed eigenbasis. The first clustering strategy is a centralized one, for which we treat the energy concentration of eigenbasis as sparsity feature vector and use traditional pattern clustering method to divide the nodes into clusters. The second one is a distributed heuristic strategy simultaneously considering residual power, communication cost, and basis energy distribution over clusters. By utilizing eigenbasis, both strategies are independent of the data to be collected and applicable in irregularly placed WSNs. Simulation results from both synthetic and real data are included to demonstrate the proposed strategies.