A Minimum Spanning Tree based Clustering Algorithm for Cloud based Large Scale Sensor Networks

Abstract

Wireless sensor networks (WSNs) can be composed of huge numbers of nodes collecting data from the environment. WSNs are crucial communication layer technologies of Internet of Things. The obtained data by the WSNs can grow exponentially, hence utilizing big data analysis techniques and cloud computing technologies are of utmost importance. WSNs can be used in various applications such as habitat monitoring, military surveillance, smart agriculture, miner safety and healthcare applications. Sensor nodes are generally battery-powered, so conserving the residual energy of nodes is very important to prolong the lifetime of the applications. WSNs do not own a fixed infrastructure, hence messages of the applications transmitted in an ad hoc manner to the sink node. Since the transmission range of sensor nodes are limited, multi-hop communication is used. Clustering is a very important method for supporting multi-hop routing in WSNs. Data aggregation, time synchronization and load balancing are some of the well-known operations that benefit from clustering. Selecting efficient communication paths and distribution of nodes evenly to partitions in clustering operation lead to boost the network lifetime. In this paper, we propose a minimum spanning tree based clustering and backbone formation algorithm (MICUB) for WSNs. The proposed algorithm inputs node coordinates, transmission range, sensing area dimensions and partition numbers and outputs clustering and backbone information. MICUB algorithm first forms a minimum spanning tree backbone and divides the networking area into equal partitions where each partition is a cluster. In this manner, efficient links are selected for backbone formation and the clusters are constructed evenly. The intra-cluster links are constructed by again executing a minimum spanning tree algorithm inside the clusters. We measure the coefficient of variations of the proposed MICUB algorithm and its counterparts to obtain the clustering quality. These results show us that our proposed algorithm performs very well against node counts and degrees.