Localization For Wireless Sensor Networks Via Norm Regularization Matrix Completion

As an important technical component of wireless sensor networks,scholars at home and abroad have always paid attention to the research of positioning technology.The popularity of wireless sensor network applications in many fields has led to a corresponding increase in the requirements of the sensor network for positioning technologies,such as positioning accuracy,noise immunity,and positioning cost.The necessary condition for most wireless sensor network applications is to obtain correct and sufficient location information.However,existing positioning algorithms often suffer from data noise pollution and data loss in the positioning process.In recent years,with the development of machine learning,matrix complementarity theories have become increasingly diversified,so a positioning algorithm based on matrix completion has emerged,but these algorithms do not consider noise interference,or only consider the interference of some common noises.The lack of consideration of the structural noise interference that may occur in practical applications,and therefore poor performance in positioning accuracy.In order to solve the problems in the above positioning,thesis conducts in-depth research on the existing matrix completion techniques.After analyzing and improving the existing positioning algorithm based on matrix completion,the noise-tolerant wireless sensor networks localization via multi-norms regularized matrix completion?LMRMC?is proposed.For the noisy and missing Euclidean distance matrix between nodes,we model it as a norm regularization matrix complement optimization problem,in which outlier noise is modeled asL₁norm and structured noise is modeled asL_1,2norm,Gaussian noise is implicitly filtered through the penalty factor during the solution process.The model considers Gaussian noise,outlier noise,and structured noise at the same time,and further introduces the stochastic proximal gradient descent?SPGD?algorithm into the LMRMC to reduce the amount of computation and storage of the positioning algorithm to save nodes.energy.The simulation results show that the proposed algorithm LMRMC can locate the noise location while achieving better node positioning accuracy.This feature provides powerful support for the diagnosis and maintenance of wireless sensor networks.