Study On WSN Multilateral Localization Algorithm Based On RSSI

Wireless sensor network(WSN) is a new data acquisition technology developed in the recent years which includes modern sensor technology,MEMS, communication technology,embedded compute technology,distributed information processing technology,etc.WSN has broad application fields such as military reconnaissance,environmental monitoring,medical monitoring and building monitoring,etc.As the important support technologies of WSN,localization becomes a key problem.Although some kinds of localization algorithms are proposed,such as Received Signal Angle Localization,DV-hop Localization,Centroid Localization and APIT,etc,low localization accuracy is widespread in the current localization algorithms as localization is an ill-posed problem in the reason of ranging error.In this paper,based on Tikhonov regularization and Tikhonov iteration regularization methods two localization algorithms are proposed for the ill-posed problem in the multilateral localization.The location model based on RSSI,the channel attenuation index n,the optimal reference node number,the regularization parameterαand the corresponding stopping rule of iteration are studied.Research results show that the best localization results can be acquired while the Log-distance Path Loss Model channel attenuation index n is 4 and the reference node number is 5 in the plane localization.The minimum localization errors can be to less than 1 meter while the Tikhonov regularization parameterαis 400,the Tikhonov iteration regularization parameterαis 700 and the correspond stopping rule of iteration is△≤3.The iteration convergence needs about 5 steps.In the spatial localization,the optimal measuring distances can be acquired while the Log-distance Path Loss Model channel attenuation index n is modified as 6 to cope with the obstruction.The localization errors can be generally reduced to about 2 meters while the Tikhonov regularization parameterαis 50,the Tikhonov iteration regularization parameterαis 100 and the correspond stopping rule of iteration is△=‖X_α,δⁱ⁺¹-X_α,δⁱ‖&△≤1.The iteration convergence needs about 5 steps.These results are significantly better than those of the Maximum Likelihood Estimation (MLE) method.