Research Of Clustering Algorithm For Wireless Sensor Networks

Wireless sensor networks combining with microsensor technology, low power computing and wireless communication technology, have become an active branch of the researches in computer science. Because clustering architecture is better than planar architecture on energy saving and network expansibility,research against clustering architecture is a hopspot now.The clustering architecture depends on some clustering algorithms,which have a direct influence on the working level and the availability of the wireless sensor networks.In this paper some classical clustering algorithms are analyzed and and their applications are discussed.First,through the researches against the algorithm based nodes'density,an improved algorithm is proposed,i.e.CAND-T.In the algorithm,the network threshold is estimated to restrict every cluster's scale and reduce overlap among clusters.Second,a clustering algorithm based RBF netural network is proposed,i.e.CARBF.In the algorithm three factors such as density, energy and location are comprehensively considered to select cluster-heads.The overall information is provided by distributed temporary clustering approach, and then clusters are formed via the base station which has enough energy.In this paper CAND-T and CARBF algorithms are described in detail and simulation experiments are made.The result demonstrates that CAND-T algorithm raises LBF(load balancing factor) and reduces overlap among clusters, which adapts to be applied in the field with a large number of nodes.And the simulation result demonstrates that CARBF algorithm can make cluster-heads well distributed,raise the network connectivity and LBF and extend the lifetime of network comparing with LEACH and LEACH-C algorithm by the simulator made by myself, which has a great use value.