Layered adaptive compression design for efficient data collection in industrial wireless sensor networks

Existing compressed sensing (CS)-based spatiotemporal data compression schemes can significantly decrease communication consumption for data collection; however, they ignore data correlation among different clusters over spatial dimensions. To explore data correlation among different clusters and satisfy the requirement of high data precision in industrial applications, in this paper, we propose a layered adaptive compression design for efficient data collection (LACD-EDC) in industrial wireless sensor networks (IWSNs). In the proposed scheme, first, we design a multilayer network architecture to support the exploration of spatiotemporal correlations, especially spatial correlation among different clusters. Then, we construct specific projection methods for exploring temporal correlation in sensory nodes, spatial correlation (intracluster) in cluster heads and spatial correlation (intercluster) in processing nodes. In addition, a detailed solution method is developed to recover the original data and achieve approximate data collection in the sink node. Subsequently, sparsifying dictionaries are trained for adapting different types of data and obtaining better sparse representations, which further improves the data recovery accuracy. Our simulation results indicate that the proposed layered adaptive compression scheme offers better recovery performance than conventional clustered compression schemes (i.e., achieving efficient data collection with high quality).