Technical report: Geometry Aware Ad-Hoc Localization

Localization is essential part of network control in many ad-hoc and sensor network applications. Many localization scenarios handle just simple, uniform and isotropic network structures and often some localization aware nodes called anchors or beacons are needed. Here we present a geometry-aware localization algorithm, what can operate in unisotropic networks that have complex topology. To alleviate error cumulating during computation, network is partitioned into a set of locally isotropic clusters. Then nodes are first localized without beacons in clusters to form local coordinate systems. When executing that computation, the effect of geometry to localization error is taken into account by using online metric Geometric Dilution of Precision (GDOP). After that an overlapping algorithm identifies a set of pivot nodes in each cluster, and then local coordinate systems are merged to get everything in one coordinate system. Error behavior is observed, and in addition to online metric GDOP, also offline metric Cramér-Rao Lower Bound (CRLB) is used in error estimation.