Accurate Range-Free Localization for Anisotropic Wireless Sensor Networks.
Shigeng Zhang,Xuan Liu,Jianxin Wang,Jiannong Cao,Geyong Min
DOI: https://doi.org/10.1145/2746343
2015-01-01
ACM Transactions on Sensor Networks
Abstract:Position information plays a pivotal role in wireless sensor network (WSN) applications and protocol/algorithm design. In recent years, range-free localization algorithms have drawn much research attention due to their low cost and applicability to large-scale WSNs. However, the application of range-free localization algorithms is restricted because of their dramatic accuracy degradation in practical anisotropic WSNs, which is mainly caused by large error of distance estimation. Distance estimation in the existing range-free algorithms usually relies on a unified per hop length (PHL) metric between nodes. But the PHL between different nodes might be greatly different in anisotropic WSNs, resulting in large error in distance estimation. We find that, although the PHL between different nodes might be greatly different, it exhibits significant locality; that is, nearby nodes share a similar PHL to anchors that know their positions in advance. Based on the locality of the PHL, a novel distance estimation approach is proposed in this article. Theoretical analyses show that the error of distance estimation in the proposed approach is only one-fourth of that in the state-of-the-art pattern-driven scheme (PDS). An anchor selection algorithm is also devised to further improve localization accuracy by mitigating the negative effects from the anchors that are poorly distributed in geometry. By combining the locality-based distance estimation and the anchor selection, a range-free localization algorithm named Selective Multilateration (SM) is proposed. Simulation results demonstrate that SM achieves localization accuracy higher than 0.3r, where r is the communication radius of nodes. Compared to the state-of-the-art solution, SM improves the distance estimation accuracy by up to 57% and improves localization accuracy by up to 52% consequently.