Self-localization in wireless sensor networks using particle filtering with progressive correction

Abstract

A centralized self-localization algorithm is used to estimate sensor locations. From the known positions of at least 3 anchor nodes the remaining sensor positions are estimated using an efficient particle filter (PF) with progressive correction. The measurement model is a simple two-parameter log-normal shadowing model, where the parameters are estimated concurrently. Experiments using Crossbow Imote2 motes show that an error of less than 16% is achievable in an indoor environment. The results demonstrate that by using PF with progressive correction, a small number of measurements and a simple signal propagation model are sufficient to give low localization errors.