Research Of Localization And Tracking Algorithms For Wireless Sensor Network Based On Transductive Regression

Wireless Sensor Network(WSN), a novel data acquisition technique, integrate multifold subjects including microelectronic, wireless communication and wireless network, and is widely used in military, industrial controlling, environmental monitoring and medical assistance fields. In most applications, determining the physical positions of sensor nodes is the basic requirements. However, a large number of sensor nodes are deployed randomly, furthermore the nodes have limited software and hardware resuouces, it is theoretical significance and practical value to design an effective localization algorithm to identify the position of each node. The main mission of this thesis is to research several localization algorithms of WSN and propose a new node localization algorithm for low non-line of sight(NLOS) error of WSN.Above all, this thesis generalizes the architectur, features, research significance and progress of WSN, simply analyzes the basic principle of node localization, classifies several common localization algorithms and summarizes the existed main problems of the current algorithms.The next, for large NLOS error in the node localization and the shortage that support vector regression(SVR) localization method reduces the error, the thesis presents a node localization algorithm based on the transductive regression about WSN, which uses beacon node positions, radio- frequency feature during nodes and kernel function to locate unknown nodes. The advantage of the algorithm is to request only test node set to get smaller error, without regard to the error of the whole sample set(including train node set).Once again, using the algorithm in the circumstance of mobile node localization, the thesis also presents a mobile localization and track algorithm based on the transductive regression, which includes two steps. The first step uses the transductive regression method to estimate the sample nodes of the mobile node at the different moment; the second step is tracking for the mobile node.The two algorithms are made simulations and result analysis, and comparison to the SVR method. The experiment results indicate that the transductive regression localization method not only reduces localization error in the static node localization situation, but also can improve the localization precision in the mobile node localization situation. In the end, it is the conclusion of the thesis and the prospect for the future research work.