Performance of angle-and-range-based localization of wireless sensors

In wireless sensor networks, the accuracy of localization information deteriorates significantly after relaying multiple hops due to accumulation of errors. Techniques exist to enhance the localization information by employing redundant measurements available in networks with high connectivity or from a large number of reference nodes. However, it is challenging to obtain accurate localization information in deployment environments with: 1) unfavorable radio-propagation characteristics that degrade measurements, or 2) sparse topologies that limit the redundancy. One such example is underground tunnels. In this paper, we study the performance of localization techniques that are applicable to such sparse networks. Particularly we focus on the COBALT method, which is based on cliques of nodes in the network and uses angle of arrival and range estimates. For different techniques, we show the influence of the connectivity among sensor nodes on the accuracy. Our results show that the COBALT method outperforms other general-purpose localization techniques.