A New Two-Level Clustering Scheme for Partitioning in Distributed Wireless Sensor Networks

Partitioning has always been a challenge in the design of distributed applications. It allows optimizing the intercommunication between the system components and so increasing the lifetime of the network. Graph theory methods have often been used to perform partitioning in classic distributed systems but seem to be not efficient in ad hoc or wireless sensor networks (WSN). The main reason is related to the topology of these kinds of networks and the presence of multihop communication. In this paper, we propose a new self-organisation of the WSN based on the optimization of the number of jumps between any sensor and the sink. The network is based on a two-level hierarchy structure and organised as a set of clusters with one cluster-head by cluster and a super-leader for the entire network. The optimisation process has been performed and validated by introducing some parameters, baptized cohesion parameters. The simulation of our approach compared to existing and previously developed protocols shows the efficiency of the method. The results are very interesting and allow projecting several perspectives to improve performances by using other metrics.