Research On WSN Mobile Node Localization Algorithms Under NLOS Environment

WSN (Wireless Sensor Network), which is an emerging information acquisition and processing technology, has broad application prospects in many fields. As one of the key technologies of WSN, the localization of mobile nodes is one of the hottest issues in researches of WSN. Current researches only focus on LOS (Line-of-sight) environment. But in practical application environments, NLOS (Non-line-of-sight) propagation phenomenon of signals is ubiquitous, and will decrease the accuracy of localization algorithms considerably. In this thesis, the node localization issue under NLOS environment is deeply discussed and researched with the view of improving the localization accuracy of mobile nodes under NLOS environment.In this thesis, the characteristics of the residual of NLOS distance measurements are analyzed in detail, and strict residual selection mechanism is proposed to identify the state of distance measurements. Since the distance measurements typically contain both LOS measurements and NLOS measurements, making full use of LOS measurements can effectively improve the localization accuracy of the algorithm. In this thesis, at first a linear regression model of EKF (Extended Kalman Filter) algorithm is utilized to generate the residuals of the measurements. Then the difference between the residuals of LOS measurements and the residuals of NLOS measurements is made use of to complete the accurate identification of the state of distance measurements. The simulation results show that under NLOS environment applying strict residual selection mechanism to identify the states of measurements and combining unfixed beacon nodes EKF algorithm to localize the mobile node can obtain higher localization accuracy.Based on the localization idea of M-Estimator algorithm, a robust mobile node localization algorithm is proposed in this thesis. Through analyzing the statistical property of NLOS residual, this thesis proposes a variable kernel density estimation algorithm based on nearest neighbor approach to estimate the probability density function of residuals. Then combined with the localization idea of M-Estimator, a mobile node localization algorithm based on variable kernel density estimator is proposed. From the results of the experiments, we can conclude that this algorithm overcomes the limitations of M-Estimator algorithm which needs model matching and artificial adjustments of the parameters, and successfully suppresses the NLOS errors under different environments.Considering the characteristics of NLOS errors, this thesis proposes a probabilistic data association algorithm based on vote selection mechanism. Employing the property that the standard deviation of NLOS errors is larger than that of measurement errors and combining the idea of high-frequency ranging data processing, this thesis proposes a data processing algorithm based on vote selection mechanism to filtrate the distance measurements and reserve the reliable measurements. Then a modified probabilistic data association algorithm is proposed to fuse the multiple measurements which are reserved from the vote selection. The location of the mobile node is finally determined by linear least squares algorithm based on reference beacon node selection. This algorithm effectively weakens various kinds of NLOS errors and largely improves the localization accuracy of the mobile node.WSN mobile node localization algorithms under NLOS environment have been systematically researched in this thesis, and they have been detailedly analyzed through a series of simulation experiments and field experiments. The good robustness and high localization accuracy of the algorithms proposed in this thesis is proved by experiment results.