Graph theory based aggregation of sensor readings in wireless sensor networks

Abstract

Two of the fundamental challenges associated with data gathering in sensor networks are data classification and data aggregation. This paper provides a solution to classify and aggregate sensor readings. We leverage our previous experience and use Competitive Learning Neural Network (CLNN) as the data classification mechanism. We then propose and evaluate Graph Theory Based Aggregation (GTBA) which combines outputs of CLNN across the network. We have evaluated two main interpretations of GTBA on real data sets produced by the WSN and on a testbed consisting of MicaZ motes. We demonstrate its ability to deduce an accurate representation of the data and distinguish the noise free data with a high probability.