A Optimized Distributed Localization Algorithm Based On Multidimensional Scaling Of Wireless Sensor Networks

Wireless sensor networks are a kind of new information processing platform equippedwith sensing, computing and communication ability. They are widely applied to many fieldssuch as national defense,environmental monitoring,traffic control and medical treatment.Nodelocalization is one of the most important technology for the wireless sensor networks.There isno actual meaning for the nodes propagate the messages without their positions.There arecentralized algorithm and distributed algorithm to be used to obtain the unknown nodes’location informations.As the centralized algorithm requires the high-energy calculationnodes,it is not very usefull for the large or irregular networks.If the computed node runs out ofthe energy,the network would be no practical significance.The distributed computingalgorithm does not require high-energy nodes,it is able to adapt to a variety of networks.The first MDS-MAP algorithm based on the classical multidimensional scaling is acentrialized location algorithm,it can be used in both range-based and range-freeenvironment,but it requires high-intensity computing nodes and there are more errors in theirregular region network.Improved MDS-MAP algorithm based on the classicalmultidimensional scaling is a distributed algorithm,it reduces the complexity of computing forthe nodes.But in range-free environment,the accuracy is not very high.The MDS-MAPalgorithm based-on non-metric multidimensional scaling is an improved algorithm for thisdistributied algorithm,it improves the calculation accuracy in the range-free environment,butevery node needs to calculate the shortest path in the regional network it belongs to.Thecomputational complexity is high and every node consumes much energy.This paper presents a distributed algorithm based on multidimensional scaling.It dividesthe network into multiple small networks and selects one cluster head node for each smallnetwork.The algorithm calculates the relative coordinates by cluster head nodes and integratesthem into total coordinate through the fusion algorithm.It converts the relative coordinates toabsolute coordinates by the informations of the beacon nodes and improves the accuracy ofthe nodes by matrix iterative optimization algorithm.The algorithm based on clusteringalgorithm reduces the computation for the nodes and the number of the local networks,itimproves the accuracy in both range-based and range-free environment,and it is more adapt to the irregular network.Based on Matlab2009a,this thesis has simulated all the models and methods above,andanalysed the results detailly.