Multi-pass maximum likelihood technique for self localisation in wireless sensor networks

An accurate self-localisation capability is highly desirable in wireless sensor networks. This paper presents a multi-pass algorithm, based on maximum likelihood optimisation to estimate the locations of sensor nodes within a network using range based measurements. Flip ambiguities are a major problem in trilateration based localisation, which has the potential to introduce large errors in the location estimates. The notion of robust quadrilaterals has been incorporated with the multi-pass algorithm to identify these flip ambiguities and provide good starting positions for the optimisation. The Crame¿r-Rao Bound of the estimation error has been derived which is used to analyse the accuracy of the algorithm. By fusing two complementary range-based measurements it is observed that the multi-pass algorithm follows the calculated bound on simulated data. Experiments using real data from MICAz motes demonstrate that the algorithm can maintain this accuracy in a real environment.