An in-network approximate data gathering algorithm exploiting spatial correlation in wireless sensor networks

Recently, several studies have proposed to utilize spatial correlation of sensor readings and approximate answers to select only a subset of representative nodes for reading reporting in the WSN monitoring applications. Although having achieved substantial energy saving, these studies suffered from high control overhead or did not fully exploit spatial correlation. In this paper, we propose an in-network approximate data gathering algorithm exploiting spatial correlation. The proposed algorithm consists of two phases: in-network clustering phase and reading streaming phase. In the former phase, we present an in-network clustering scheme exploiting spatial correlations of sensor readings as well as cluster readings to further reduce the number of representative nodes. On the other hand, the latter phase employs an adaptive cluster maintenance scheme that ensures the user to obtain the reading answers of desired quality despite changing sensor readings. The experimental results show that the proposed algorithm outperforms prior algorithms in terms of network lifetime and number of representative nodes.