Two-phase clustering based aggregation of sensor data (Q601265)

Summary: To conserve energy, wireless sensor networks need to send only interesting data to the base station. Since the general nature of the data to be sensed is known a priori, the present work proposes ART based clustering and compares it with density based clustering of sensed data. ART based clustering works in two phases, offline and online. Fuzzy ART is employed as initial clustering process in the pre-deployment offline phase to detect natural clusters in data. To cater to the dynamic nature of the environment, fuzzy ARTMAP neural network (FAMNN) is used for clustering after deployment in the online phase. The second technique density based aggregation (DBA) is based on density of points within the clusters. Simulation results on synthetic data indicate that FAMNN is suited for wireless sensor networks. DBA is able to match the initial clustering ability but creates large number of outliers as seen with the data under test.