Cluster-Based Energy-Efficient Data Collection in Wireless Sensor Networks Utilizing Compressive Sensing

In this paper, an integration of compressive sensing (CS) and clustering in wireless sensor networks (WSNs) is proposed to significantly reduce the energy consumption related to data collection in such networks. Both compressive sensing (CS) and clustering have been proved to be efficient ways to reduce the energy consumptions in WSNs, however, there is little study about the integration of them for further gains. The idea is to partition a WSN into clusters, in which each cluster head collects the sensor readings within its cluster and forms CS measurements to be forwarded to the base station. The spatial correlation of the readings in a WSN results in an inherent sparsity of data in a proper basis such as discrete cosine transform (DCT) or Wavelet. This sparsity can then facilitate the application of the CS in data collection in WSNs. This way, we only need to forward « N CS measurements from N sensor nodes. An important issue that needs to be considered for applying CS in the data collection problem is the underlying routing mechanism. Some related studies employ minimum spanning tree, random walk, or gossiping as the routing mechanism. However, we propose applying CS on top of a clustering algorithm to reduce the energy consumption. Under this novel framework, we study different clustering techniques and the properties of the block diagonal measurement matrix that is formed based on the clustering algorithm. We further formulate and analyze the total power consumption, based on that we can obtain the optimal number of clusters for reaching the minimum power consumption.