The Research On WSNs Localization And Node Optimized Selection Technologies In Sparse Application Scenarios

As an important part of the Internet of Things(IoT),Wireless Sensor Networks(WSNs)has the unique advantages of advanced technology idea,fast and flexible networking,convenient deployment,etc.WSNs can extend the existing network to the physical world,and change the way of human interaction with the physical world.In recent years,it has been highly concerned by researchers and industrial enterprises in the world.At present,WSNs has been widely used in many fields,such as battlefield monitoring,geological disaster relief,logistics transportation,intelligent home,environmental protection supervision and so on.However,since the WSNs node energy,communication range,computing and storage capacity are limited,there are still many significant technical challenges in the face of continuous growth of complex scene mode and diversification of the specific business needs.To save the network energy and prolong the life cycle of network,in this paper,we utilize the theory of sparse signal processing to solve the problems of WSNs target localization,node localization and node optimized selection in sparse application scenarios.The main research work and innovation are listed as follows:(1)A novel target localization algorithm based on improved orthogonal matching pursuit technique is proposed.Aiming at the sparse characteristics of WSNs target localization problem,we first divide the localization area into grids,and deploy multiple sensor nodes to measure the received signal strength of target nodes.In such case,the target localization problem is converted to sparse search problem,and the localization model based on compressed sensing is established.Then,the orthogonalization preprocessing method is adopted to make the measurement matrix better satisfy the restricted isometry property.Finally,from the perspective of location recovery,we improve the orthogonal matching pursuit technology by selecting multiple atoms in each iteration,which reduces the computational complexity theoretically,and has superior anti-noise ability.(2)We propose a multi-target localization algorithm based on sparse Bayesian learning.To satisfy the requirement of the restricted isometry property of the measurement matrix for compressed sensing theory,a matrix decomposition method which does not affect the sparsity of the original signal is applied to preprocess the measurement matrix.In order to further improve the accuracy of multi-target localization,we utilize sparse Bayesian learning based on variational approximation,which can express the sparse signal recovery as a linear regression problem,and approximate the sparse prior knowledge to joint posterior distribution parameter of the model.To solve the error caused by gridding assumption,an approach based on K-means clustering and weighted centroid localization is proposed to localize the targets that are not located at the center of grid.(3)We propose a node location algorithm based on Euclidean distance matrix completion.Firstly,the node localization problem based on the incomplete Euclidean distance matrix is equivalent to the low-rank matrix completion problem with a group of linear equality constraints.Then,based on the low-rank property of the Euclidean distance matrix and the matrix factorization,a matrix completion method based on sparse Bayesian learning is proposed to restore the original Euclidean distance matrix.Finally,the relative positions between nodes are obtained by multi-dimensional scaling technique,and then are further transformed into the absolute positions by using the coordinate information of the anchor nodes.The approach uses only a small amount of acquisition data to localize nodes with high precision,which improves the node localization efficiency,and is especially suitable for large-scale wireless sensor networks.(4)An optimized sensor node selection algorithm for sparse estimation is proposed.We adopt the lower bound of mean square error of the minimum mean square error(MMSE)estimator as the objective function for estimating the WSNs sparse signal,and propose optimized sensor node selection algorithms in the MMSE sense under correlated and uncorrelated noise cases.To solve the non-convexity of the optimization problem,a convex relaxation method is proposed,which transforms the node optimization problem into a suboptimal semi-definite programming problem.Furthermore,the weighted relaxation algorithm is used to solve the problem of fractional solution caused by Boolean constraint relaxation.At last,an equivalent stochastic optimization method is proposed to solve the problem of high computational complexity.