Research On Wireless Sensor Network Localization Algorithms And Their Applications

The wireless sensor network (WSN) is a self-organizing network consisting of a largenumber of sensor nodes with sensing, computing and communicating functions. The WSNcan integrate the logical information world and the real physical world together, and has wideapplications in military and civil areas. In many WSN applications, the location informationof sensor nodes is very important in the lifetime of the entire network. Designing nodelocalization algorithms of high-efficiency, high-precision and low power consumption hasrecently been one of the most concerned research topics of the WSN.This thesis is mainly focused on certain practical problems about current nodelocalization algorithms used in WSN applications. Using the dimension reduction technologyas the basic research tool and aiming at decreasing the localization error and increasing theapplicability of algorithms, we perform in-depth study on node localization techniques fromthe viewpoints of the anchor node ratio, the communication distance and the density of sensornodes, the ranging error, and the location accuracy. We propose three types of novellocalization algorithms with both theoretical and practical values based on the matrix analysistechnique, the principal manifold algorithm and the maximum likelihood estimate algorithm.Furthermore, we integrate the WSN with the traditional video monitoring system to realize anintelligent multiple targets tracking video system. The main research content andcontributions of this thesis are summarized as below.1. Relative researches on the node localization algorithms for WSN are firstly reviewed.Comprehensive analyses are performed on current localization algorithms by consideringparameters such as the anchor node ratio, the communication distance and the density ofsensor nodes, the ranging error and the location accuracy. The drawbacks and unsolvedproblems of existing localization methods in practical applications are then summarized.2. After analyzing the disadvantages of the centralized multidimensional localizationalgorithms MDS-MAP (Multi-Dimensional Scaling-MAP) in positioning accuracy andcomputational complexity, we present a new localization algorithm based on a set ofuncorrelated discriminant vectors (SUV). The solving equation of the double centered matrixcan be simplified by node coordinate transformation. In order to reduce the noise disturbanceand decrease the effect of ranging error on the followed location accuracy, a new coordinateinner product matrix can be reconstructed by using a set of uncorrelated discriminant vectors,which can be used to calculate the node coordinates directly. This algorithm can realizecentralized localization, distributed localization and incremental localization of nodes.3. After studying the topological structure of neighboring nodes in the WSN, we presenta local subspace alignment (LSA) algorithm by combining with the idea of the principal manifold learning and nonlinear dimension algorithm. This algorithm is particularly suitablefor determining the relative locations of sensor nodes in the large-scale and low-densityWSNs, where the low connectivity between nodes and the large ranging error betweenlong-distance nodes usually make accurate localization quite difficult. In this algorithm, basedon the pair-wise distance between each node and its neighbour nodes within a certaincommunication range, the local geometry of the global structure is firstly obtained byconstructing a local subspace for each node, and those subspaces are then aligned to give theinternal global coordinates of all nodes. Combined with the global structure and the anchornode information, we can finally calculate the absolute coordinates of all unknown nodes bythe least squares algorithm.4. The ranging error of WSN is usually large in complex environments. We find that theelements of the coordinate inner product matrix may fluctuate in a certain range with thechanging ranging error. So we present a maximum likelihood estimate algorithm based on thecoordinate inner product matrix for determining the relative locations of sensor nodes incomplex environments with large ranging error. Based on the global topological structure andthe connectivity of WSNs, the geodesic distance between each node and the coordinate innerproduct matrix are obtained. Using the maximum-likelihood estimator for coordinate innerproduct matrix, we can finally estimate the sensor node coordinates by finding the globaloptimal solution.5. WSNs have advantages in the event monitoring, and the traditional video monitoringsystem is usually not intelligent enough. So we integrate the two systems to realize anintelligent indoor multi-target tracking system, which can effectively detect, locate and tracktargets in the monitored area. We present a solution combining the WSNs with the Ethernettechnology and design the routing method to decrease the received signal strength indication(RSSI) interference by buildings. The implemented system can thus perform continuouslocalization and tracking in different rooms or different floors in a building. Since the layoutsof the room and the corridor are different, we design the nearest neighbor algorithm fortracking targets in corridors, and use the distributed SUV algorithm in rooms. Finally, sincewireless signals are usually influenced by different noises in the indoor environment, wedesign a RSSI smoothing method to increase the localization accuracy by removing noisesand abrupt changes from the received RSSI values before using them to calculate the distance.