Distributed Iterative Node Localization via an Advanced K-Means++ Procedure in WSNs

Distributed localization is an attractive approach for large-scale wireless sensor networks (WSNs). However, conventional distributed localization methods suffer from several issues, such as dividing the WSN into nonoverlapping subregions and resorting to complicated algorithms such as semi-definite programming (SDP) or second-order cone programming (SOCP). These methods often lead to lower localization accuracy for nodes far from the subregion center. To overcome these challenges, this article proposes a distributed iterative node localization method that uses an advanced
${K}$
-Means++ partition algorithm to divide the WSN into multiple overlapping subregions. Within each subregion, an iterative approximate algorithm is used to solve the local localization problem, which is much more efficient than SDP and SOCP. Finally, a weighted fusion strategy is designed to refine the positioning of the nodes in the overlapping subregions. Numerical simulation results demonstrate that the proposed method achieves higher and more robust localization accuracy in large-scale WSNs compared with existing methods.