View article

Localisation is required for many ad-hoc sensor network applications. Therefore, much work has been done regarding techniques for localisation, mainly using anchors (nodes with known locations). However, there has been little study of how anchors are likely to be distributed in applications, and how to perform localisation with more realistic anchor distributions. In this paper we look at the limitations of many of the existing proposed localisation techniques with regards to coping with non-uniform anchor distributions and errors in ranging information. We present a refined approach using mobile anchor scenarios for anchor information distribution, combined with statistical techniques for performing localisation with inaccurate range data. We also show methods for dealing with motion in nodes using related techniques, including an anchor-less solution to ensure that we can always detect motion. Simulations with our refined approach have shown significant reductions (in the order of magnitude range) to the required processing for performing statistical localisation over previous attempts, as well as improving the generated location information in situations with non-total anchor information coverage, making possible a wider range of applications.