Spatiotemporal correlation–based adaptive sampling algorithm for clustered wireless sensor networks

Abstract

Energy efficiency is one of the most crucial concerns for WSNs, and almost all researches assume that the process for data transmission consumes more energy than that of data collection. However, a few sophisticated collection processes of sensory data will consume much more energy than traditional transmission processes such as image and video acquisitions. Given this hypothesis, this article proposed an adaptive sampling algorithm based on temporal and spatial correlation of sensory data for clustered WSNs. First, according to spatial correlations between sensor nodes, a distributed clustering mechanism based on data gradient and residual energy level is proposed, and the whole network is divided into several independent clusters. Afterwards, each cluster head maintains an autoregressive prediction model for sensory data, which is derived from historical data in the temporal domain. With that, each cluster head has the ability of self-adjusting temporal sampling intervals within each cluster. Consequently, redundant data transmission is reduced by adjusting temporal sampling frequency while ensuring desired prediction accuracy. Finally, several distinct sampler collection sets are selected within each cluster following intra-cluster correlation matrix, and only one sampler collection needs to be activated at each round time. Sensory data of non-sampler can be substituted by those of sampler due to strong spatial correlation between them. Simulation results demonstrate the performance benefits of proposed algorithm.