Study Of Three-dimensional Localization Method Based On Multidimensional Scaling In Wireless Sensor Networks

It is indispensable to position nodes of wireless sensor networks for applications, in most situations, nodes are randomly deployed in the monitoring areas which is unpredictable and usually is in three-dimensional space. Therefore, it is necessary to position the nodes in three-dimensional space. This paper is sponsored by Jiangxi Scientific and technological project which studies on three-dimensional localization algorithm specific to the features of WSN. It includes how to reduce the costs of networks and refine the accuracy of position consider a large amount of nodes with low capability of computation, complex of topology and restrict in energy.The paper proposes a three-dimensional Localization method based on Multidimensional scaling for Wireless Sensor Networks, named 3D-MDS for short. Through studied on the technique of Multidimensional Scaling and the method of range measurement specific to the large-scale WSN, it proposes using topological control method and global coordinates system to divide the network in logical. The amount of nodes in each cluster is smaller for using of cluster structure, as well as the amount of multi-path range between nodes. This can be assured that the scale of dissmilitery matrix and the range error is under controlled. For reducing range measurement error, the paper utilizes RSS experience attenuation model to measure one hop-distance between nodes and discuss on using kalman flitter to reduce the noise. Moreover, to the characteristic of MDS, a range correction method which combines the variable power levels and the triangle principle is proposed. For each cluster-head it applies a lightweight eigen-decomposition method to compute the three-dimensional coordinates for decreasing the complex of computation. Then, each cluster-head picks a certain number of in-cluster nodes to establish the global coordinates and finds a shortest path between nodes which are in different cluster through the backbone, and estimates the range by accumulate the one-hop-distances in the path. The global coordinates and the transform variables are computed by Sink or background and send back to the cluster head in where to finish the in-cluster global coordinate transformation. Due to the low accuracy of light-weight eigen-decomposition method, the background would to refine the coordinates of in-cluster nodes using an iterative optimization algorithm finally. In the final of this paper, a position test bed based on the 3D-MDS localization algorithm is designed. The mainly function is finished under the platform of TinyOS 2.0 and TelosB, as well as some experiments.