Skip to Main Content
Clustering is a fundamental mechanism to design scalable protocols for sensor networks. In this paper we advocate to use this clustering functionality to perform data aggregation and network formation. We investigate scenarios of event-driven sensor networks with high degree of spatial-temporal correlation. Our main focus relies on the cross-layered design and evaluation of localized algorithms for performing rapid data aggregation and network formation allowing prompt response to queries. The localized algorithms are based on probabilistic clusterhead election. The arbitration in gathering the data from the cluster members is based on a splitting tree algorithm. Through these algorithms, each sensor node participates in the data aggregation. On the other hand, to decide whether to participate further in the ad-hoc networking infrastructure, a node applies local criterion. To minimize the energy consumption, a node goes to sleep in case it decides not to participate in the networking functionalities. We closely investigate the impact of these local algorithms and criterions to the global properties of the created sensor network.