Abstract:

The node localization problem in large scale wireless sensor networks can be formulated into a highly nonlinear nonconvex optimization problem which is hard to solve directly in large scale sensor networks.This paper proposes a new distributed localization algorithm to solve this problem.First,the global undirected graph composed of the large scale wireless sensor network is decomposed into a series of partially overlapping subgraphs,and then the global optimization problem is decomposed into a series of small scale subproblems for iterative solutions.The optimization problem in each subgraph can be solved iteratively independently.The new distributed localization algorithm for sensor nodes consists of two steps in each iteration,First,the Barzilai-Borwein gradient method is used to estimate the location of the node in the divided partially overlapping subgraph.The gradient method has a low computational cost and greatly speeds up the convergence.Second,the same sensor nodes in different partially overlapping subgraphs are fused and averaged.Theoretical analysis and simulation results show that compared with the existing methods,the proposed new distributed localization algorithm has a higher scalability and localization accuracy in large scale wireless sensor networks,and can be used for localization in large scale sensor networks.

Key words: wireless sensor networks, localization, distributed algorithm, graph model, Barzilai-Borwein gradient method