Research On Localization Of Wireless Sensor Networks

By the end of twentieth century, with progress in wireless communications, digital electronics and micro-electro-mechanical systems, wireless sensor network technologies have been developed rapidly. Wireless sensor networks, consisting of a myriad of cheap, smart sensors deployed in special environments and networked through wireless links and Internet, provide unprecedented opportunities of applications, including environments monitoring, health monitoring, vehicle tracking, inventory management and military applications. Many communication protocols proposed for traditional wireless cellular system and ad hoc networks are not well fitting for wireless sensor networks with unique features of ultra-low energy consumption and infrastructureless architecture. These features present many challenges for wireless sensor networks.Generally, measurement data is collectd by sensors and delivered to center node or processor to analyse. The collected data are meaningless without corresponding position information. The position information in collected data is very correlated with node position. And node location readings is of benefit to routing, coverage, topology control and boundary detection. Localization is an important supporting technology in wireless sensor networks. The work of this dissertation was focused on localization technology of wireless sensor networks. The-state-of-the-art of range-based and range-free localization methods was introduced. The difficulty of localization technology was analysed. Then some new positioning algorithms were introduced, including distance controlled refinement algorithms, fine-grained distance controlled localization algorithms based on updated location evaluation system and dynamic hop distances based localization algorithms.In range-based localization, range error problem and sparse anchor problem were main challenges behind positioning techniques in wireless sensor networks. The Hop-terrain algorithm proposed by C. Savarese can deal with above two obstacles, but the fraction of located sensors of the Hop-terrain algorithm is low, especially in networks with low connectivity. This paper presented a distance controlled refinement algorithm which improved upon the Hop-terrain algorithm by using the last iteration result as current reslut in some cases of iterative process. The new algorithm preserved the better interim result, improved the ability of convergence, and increased the fraction of located sensors and positioning accuracy.In range-based iterative positioning techniques, this paper analysed the importance of location evaluation system to positioning accuracy and pointed out that it is almost impossible to give a location evaluation sytem to exactly reflect positioning accuracy without knowing positions of unknown nodes. The paper analysed the change of positioning accuracy along with the number of reference nodes if there was location error of reference nodes or range error to reference nodes. Simulation results show that positioning accuracy will be increased if increasing the number of reference nodes reasonably. But the positioning accuracy will be decreased probably if increasing the number of reference nodes more. This paper presented an iterative localization algorithm based on updated location evaluation system. The new algorithm chose possible position with great accuracy as iteration result based on the constraints of distance to those neighbors with high confidence weight instead of to all of neighbors and the positioning error was decreased.In range-free localization technologies, it was difficult to decrease the area of possible region of unknown nodes in positioning methods based on area of radio coverage. It was difficult too to estimate distance between different nodes in localization approaches based on hop counts techniques, especially in networks with holes. In the paper, a range-free localization method based on dynamic hop distances was proposed which can work in anisotropic networks with holes. First, the algorithm estimated different average distances of one hop along the shortest path between different anchor pairs. Then some unknown node picked up the anchor pairs, whose average distance of one hop along the shortest path is larger than some threshold and the shortest path passes corresponding unknown node, as reference nodes to estimate its position. The new algorithm can filter out the anchors with higher errors of estimation distances as reference nodes. Simulation results show that the localization accuracy of new algorithm is higher than previous algorithms.The achievements of the dissertation can be used to promote the development of localization technologies and the applications in wireless sensor networks.