Research On A Localization Algorithm In WSN Based On MDS

In recent years, the development of wireless sensor network has received increasedattention in many countries.Sensor network has the characteristics of low-cost, less energyconsumption and wide distribution, which make it successfully used in all areas ofpeople’s lives, like military, health care,agriculture, transportation and household items.However, the studies of wireless sensor network is normally meaningless without nodelocalization problems,therefore, wireless sensor network positioning technology hasbecome one of the key issues in the research field of WSN(Wireless Sensor Network).Firstly, the classical methods ranging and localization algorithms which are widelyused in WSN positioning technology are summarized in this paper, the advantages anddisadvantages of various algorithms are also analyzed and compared based on the study ofthe basic structure of wireless sensor network. Multidimensional scaling techniques(MDS)and MDS-MAP location algorithm which uses multidimensional scaling techniquessuccessfully are deeply studied in this paper, including both advantages and disadvantages.For the lack of MDS-MAP algorithm, a distributed localization algorithm is proposedcalled NMDS-TDOA(D) algorithm. This algorithm’s process is detailed argument, finally,the feasibility and advantages of this improved algorithm is verified by simulation. In thispaper, the main work and innovations are as follows:1.MDS-MAP algorithm based on non-metric multidimensional scaling localizationalgorithm is deeply studied. This positioning algorithm makes full use of relevantinformation between nodes, this correlation is converted into a dissimilarity matrix toobtain the coordinates of the nodes in multidimensional space, though, MDS-MAPalgorithm has a larger cost for the central node, a higher algorithm complexity, a highererror cased by shortest path algorithm with a low node density and a low efficiency formobile nodes as a centralized algorithm;2.This paper presents a distributed non-metric multidimensional scaling localizationalgorithm based on energy. Nodes in the network calculate their own residual energy basedon network model and energy model. Each node compares the average residual energy ofneighbor nodes with its own to decide whether become a cluster head, this clusteringapproach reduces the complexity of the iterative algorithm, the most critical, in existingdistributed algorithms, each node usually clusters with nodes in m hops. Compare withexisting clustering methods, the proposed clustering approach reduces the number ofcluster heads, reducing the accumulation of errors due to inter-cluster convergence, whileensuring a degree of overlap between clusters which makes the fusion algorithm a higher accuracy. In addition, this clustering way extends the lifetime of the network;3.An improved centralized location multidimensional scaling algorithm is proposed inthis paper to realize positioning the cluster nodes by cluster head node. Multidimensionalscaling algorithm obtains the dissimilarity matrix between nodes by the shortest pathalgorithm,but when there is a low node density or an irregular network topology, theshortest path algorithm’s error is big. In this study, nodes will be divided into two kinds,one could use weighted geometric distance correction algorithm to correct the shortest pathdistance, the other obtain the correction factor by the neighbor nodes which have correctedtheir distance, in order to get estimated value of uncorrected path.Use Matlab to simulate the algorithm proposed in this paper. First, every improvement issimulated to verify the feasibility, then, the entire algorithm is integrated, analyzed andcompared with the classical algorithms. The improved algorithm’s superiority could beseen by comparison of the simulation results.