Multidimensional Scaling-based Sensor Localization Algorithm In Wireless Sensor Networks

Wireless sensor networks can be used in various areas, ranging from military application to people tracking and environment monitoring. Localization is a fundermental issue of many wireless sensor network applications.In this dissertation, Multidimensional Scaling (MDS) sensor localization technique in wireless sensor networks is studied. First, the sensor localization algorithms, which are used to determine sensors'positions, are examined. Most existing sensor localization methods suffer from various location estimation errors that result from ranging errors, complex network topologies and anisotropic terrain. The characteristics of MDS are explored. MDS, which is a data analysis technique, has been proved to be an efficient localization scheme.However, present MDS localization algorithms including metric-MDS and nonmetric-MDS, without any exception, make use of shortest path algorithm in network to construct distance matrix. In that way, it can undoubtedly deteriorate the locating performance of algorithm in condition of anisotropic topology and low connectivity level. In this paper, we study the matrix approximation problem and its appliccation in wireless sensor networks and introduce a novel nonmetric localization algorithm, called NMDS-LRA, which based on matrix approximation. The algorithm runs non-metric MDS on the matrix which constructed by matrix approximation, it can effectively avoid the negative impact of distance error and anisotropic topology on localization performance.Based on NMDS-LRA, the NMDS-LRA(M) algorithm, used in mobile sensor networks, is presented. The current localization methods depend on the connectivity of the network, which leads to poor performance in sparse networks. In the proposed approach, more information are available by adding virtual nodes during the movement.We compared our work with MDS-MAP(P,O) and MA-MDS-MAP(P) via simulation to show that our work can promote localization precision effectively, and most importantly, performs well on range error and anisotropic topology.