The Geo-n localization algorithm

Abstract

We introduce Geo-n, a highly precise, distancebased, general-purpose localization algorithm for use in cluttered and indoor environments, where the estimated distances between an unlocalized node and anchor nodes may contain large errors due to NLOS signal propagation. Geo-n is very resilient to outliers and to a wide range of geometrical constellations of the nodes. Geo-n uses a two stage filtering technique to obtain the most representative intersection points between every pair of circles induced by anchor coordinates and distance measurements and uses these to estimate the position of unlocalized nodes. If no intersection exists between two circles, Geo-n approximates one to further improve localization accuracy. We compare Geo-n to LLS, NLLS, AML, Min-Max, and ICLA using simulations and real world experiments, and show analysis of the spatial distribution of position errors. Results demonstrate that Geo-n outperforms the other algorithms. Like NLLS, its spatial error distribution is very homogeneous. However, Geo-n is much more robust and achieves lower average position error while retaining reasonable computational complexity.