Research Of The Node Moving Strategies In WSN

With the development of the wireless sensor networks(WSNs),recent research hotspots are the mobile sensor networks(MSNs),in which the nodes deployed are movable.The MSNs are more suitable for the applications in real world,such as the battlefield surveillance,detection of rivers,intelligent transportation and so on.Compared with traditional static networks,the most important traits of MSNs are the moving nodes.The moving nodes simplify some complex issues,for instance,the network connectivity,energy consumption,target tracking and so on.In this thesis,we mainly focus on the node moving strategies for improving the performance of target tracking.Some key points in the thesis are as follows:1.In this thesis,we build a model of heterogeneous network in which all the nodes are considered different.They are varied in three ways: the remaining energy,mobility and target recognition performance.Moreover,each property is represented by a variable ranged in(0,10),so the node in the network can be represented by a three-dimension vector.2.Based on the heterogeneous WSN model,a node moving model is proposed to improve the target tracking performance.The moving model assumes that all the nodes in the network have the same sensing range and the target can be detected by a node only when it is in the sensing range of the node.A localization scheme similar to the Centroid algorithm is utilized in the thesis.In the node moving model,some nodes which can't sense the target are chosen to move towards it.Also,the Cauchy Inequality and Induction Theory are used to prove that the moving model can improve the performance of target tracking when the distance between each moving node and the target is smaller than the localization error of the static network.3.According to the node moving model,a node moving strategy based on the genetic fuzzy tree system is presented.This thesis gives the structure of the fuzzy tree system and explain the two fuzzy inference systems constructing the fuzzy tree in detail.To tune the rule base and data base of the fuzzy tree system,the genetic algorithm is utilized.The Pittsburgh algorithm is used to design the coding of the individuals,the fitness function and the genetic operators.During the simulation process,the proposed moving strategy is firstly used in the application of locating the moving target in both two-dimension and three-dimension space.In this part,how the number of the nodes deployed in the network,the number of the moving nodes and the sensing range of the nodes influence the performance of the moving strategy is investigated.Then,the moving strategy is applied in the application of target tracking.The target is assumed to move in a line,a square or a circle,and this thesis studies the relationships between performance of the moving strategy and some network-wide parameters.The simulation results show that the moving strategy based on the genetic fuzzy tree system can indeed improve the performance of target tracking.4.In the application of target tracking,another moving strategy is proposed to make a contrast with the former one.The newly proposed moving strategy is based on the BP neural networks.Simulation results show that both of the two schemes can improve the target tracking performance in WSN.However,the genetic fuzzy tree algorithm performs better than the other one.It has lower tracking error.On the other hand,the genetic fuzzy tree algorithm needs much more computation cost than the strategy based on BP neural networks.