Sparsest Random Sampling for Cluster-Based Compressive Data Gathering in Wireless Sensor Networks

Abstract

Compressive data gathering (CDG) has been recognized as a promising technique to collect sensory data in wireless sensor networks (WSNs) with reduced energy cost and better traffic load balancing. Besides, clustering is often integrated into CDG to further facilitate the network performance. However, existing cluster-based CDG methods generally require a large number of sensor nodes to participate in each compressive sensing (CS) measurement gathering and rarely consider possible node failures due to power depletion or malicious attacks, leading to insufficient energy efficiency and poor system robustness. In this paper, we propose a sparsest random sampling scheme for cluster-based CDG (SRS-CCDG) in WSNs to achieve energy efficient and robust data collection. Specifically, sensor nodes are organized into clusters. In each round of data gathering, a random subset of sensor nodes sense the monitored field and transmit their measurements to the corresponding cluster heads (CHs). Then, each CH transmits the data gathered within its cluster to the sink. In SRS-CCDG, each sensor reading is regarded as one CS measurement, and both intra-cluster and inter-cluster data transmissions can be realized by two methods, i.e., relaying or direct transmission. Furthermore, we propose analytical models that study the relationship between the size of clusters and the energy cost when using different intra-cluster and inter-cluster transmission schemes, aimed at finding the optimal size of clusters and transmission schemes that could lead to minimum energy cost. Then, we present a centralized clustering algorithm based on the theoretical analysis. Finally, we investigate the robustness of signal recovery performance of SRS-CCDG when node failures happen. Extensive simulations demonstrate that SRS-CCDG can significantly reduce the energy cost and improve the system robustness to node failures.