Locating the Nodes from Incomplete Euclidean Distance Matrix Using Bayesian Learning

Abstract

Node localization in wireless sensor networks (WSNs) has received a considerable amount of attention. In this paper, using the natural low-rank properties of the Euclidean distance matrix (EDM), we formulate the node location finding problem from only a small fraction of random entries of the EDM as a low-rank matrix recovery problem. A Bayesian-learning-based method is utilized to recover the original EDM, based on which the relative positions of all the sensor nodes in WSNs are accurately estimated by applying classical multi-dimensional scaling (MDS). In addition, with the location knowledge of anchor nodes, we transform the relative positions into absolute positions. The simulation results illustrate that our proposed approach leads to superior performance over various other methods.