Energy efficient data gathering using prediction-based filtering in wireless sensor networks

Abstract

In wireless sensor network, sensor readings generated by nearby nodes are redundant and highly correlated, both in space and time domains. Since transmitting redundant and highly correlated data incurs a huge waste of energy and bandwidth, spatial and temporal correlation should be exploited in order to reduce redundant data transmission. In this paper, we propose an energy efficient data gathering protocol that uses a prediction-based filtering EEDGPF mechanism to solve the problem of redundant data transmissions. Our data gathering protocol organises a WSN into clusters, using data similarity that exists in readings of sensor nodes and cluster heads and uses a GARCH 1, 1 model-based non-linear predictor to exploit the temporal correlation of sensor readings. Experimental results over real dataset show that our protocol significantly outperforms linear predictor AR3-based protocol proposed in Jiang et al. 2011, in terms of number of data packets delivered, number of successful predictions and average energy consumption.