An energy efficient weighted clustering algorithm in heterogeneous wireless sensor networks

Wireless sensor networks are most used to monitor remote environments. Multitudinous sensor nodes gather data in a self-governing manner, operating on an exhaustive source of energy or battery. Clustering process structures the network into a hierarchy wherein sensor nodes gather data passed to selected cluster head nodes which perform data processing, aggregation, and transfer it to a base station. Prolonging network lifetime and enhancing total data transmission to base station are major challenges in wireless sensor network and same is addressed in this work. A distributed energy-based epoch is used in this paper to determine node eligibility to become cluster head and a multi-parameter-weighted scalarization function is proposed to determine best cluster head candidates in order to manage dynamic and multi-characteristic node heterogeneity. The parameters used are distance to base station, expected cluster head lifetime, average cluster member node lifetime and maximum power consumed by a cluster member node. A novel weight computation strategy using analytical hierarchy process is introduced in this paper which enhances the optimality of scalarization function value. The proposed algorithm is distributed over two phases as network setup phase and clustering phase. The network setup phase computes the energy model and optimal number of cluster heads. The second phase proposes the cluster head selection process using weight-based scalarization and introduces the novel weight selection method. Finally, network operation enters the data transmission phase. The results show an enhancement in throughput at base station, with an increase of close to 30% along with an increase in the network lifetime of up to 20% as measured by last node death. The simulation results are produced in comparison with the considered base protocol of DEEC as well as other protocols using similar concepts for implementation. However, utilization of a two-step cluster heads selection process including unique node epochs for shortlisting and scalarization function-based node fitness, along with optimal weight selection procedure, has led the proposed model to give better results on simulation and analyzation than preexisting algorithms.