Reserch On Nodes Localization Of Wireless Sensor Network Based On Multidimensional Scaling

Wireless sensor networks (WSN), which consist of a large number of ubiquitous sensors, is a self-organizing network system. The sensor with communication and computing ability connet each other in multi-hop wireless communicaiton way. WSN is a new informaiton acquisiton and processing technology, and have broad application prospects in military, environmental monitoring,disaster relief and business fields etc.Node localization is one of the key technologies of the application of WSN. Due to the large number of wireless sensor network nodes, and its restrictions on resources and energy, location information of nodes is very difficult to know. Therefore, the study of effective algorithm for promoting the application of wireless sensor network technology and development have the significant meaning.Multidimensional scaling (MDS), adapted from psychometric, is general data analysis technology amd have widely used in many fields. Since Shang Y. et al proposed MDS-MAP algorithm in 2003, MDS technique is used to solve node localization problem of wireless sensor networks, and have has made great progress. Connectivity information and distances between nodes are used to construct dissimilarity matrix. Relative coordinates is estimated by minimizing stress function through s series tranformation.Centralized algorithm requires strictly for sensor itself in WSN and MDS-MAP algorithm can localize accurately only if the accurate ranging. To solve these problem, an nonmeric MDS-basesd distributd algorithm, NMDS-AC, is proposed. In NMDS-AC localization algorithm, WSN is clustered to several sub-networks by selecting anchor node as cluster head. Through gathering radio Received Signal Strength Indication (RSSI) of pairwise nodes, dissimilarity matrix is constructed. Then, relative map of nodes is computed by NMDS in a cluster. All relative maps are merged into a general map. In this way, all coordinates of unknown nodes is estimated. Compare to Centralized NMDS-AC , NMDS-AC(O) reduces the requirment of computing ability of nodes. Besides, impact of ranging error is effectively reduced and computational complexity and localization accuracy is improved.For positioning error problem with error accumulation when the network size is large, an improved algorithm, MDS-AC(O), is proposed by using constant modulus algorithm of signal processing area. The proposed algorithm using NMDS-AC algorithm to estimate coordinates of unknown node, then selected as the initial value of CMA to iteratively optimize the coordinates. Positioning accuracy is improved in this way. Compared to the typical exsisting location algorithm, NMDS-AC(O) is more robuster to the ranging error and gains higher positioning accuracy. Thus it is more suitable for WSN with large-scale deployment.Thesis focuses on sensor localizaiton in wireless sensor networks. Two localization algorithms based on multidimensional scaling, NMDS-AC and NMDS-AC(O) are proposed. Simulation result show that the proposed algorithms are valid. Further research would be conducted to improve localization algorithm complex performance practicality.