Research On Compressive Data Gathering Method For Wireless Sensor Network

The Wireless Sensor Network(WSN)is a multi-hop self-organizing network system formed by a large number of micro sensor nodes deployed in the monitoring area through wireless communication.It can easily collect local signals in various complex environments and send the data to the sink or the base station for real-time monitoring of the site.The wireless sensor network is the cornerstone of the future Internet of Things.The data collection is the primary goal of wireless sensor networks.Because the data collected by sensor networks has spatio-temporal correlation,and the sensor nodes have limited resources and the sink node has powerful performance,it is very suitable for the application of Compressed Sensing(CS)theory.Therefore,the research on data collection methods for the WSN based on compressed sensing has important theoretical value and practical significance.The Compressive Data Gathering(CDG)for the WSN is based on the CS theory.Currently,the general method of the CDG is to reconstruct the original signal by collecting M weighted measurement values from M randomly generated projection nodes.The problems of this method are: Firstly,therandomness of the projection nodes selection makes the distribution of projection nodes in the network uneven,resulting in an uncertain distances between the sensor nodes and the corresponding projection nodes,and between the projection nodes and the sink,which causes the entire network unbalanced energy consumption.Secondly,collecting data through the projection node increases the energy consumption of the system,including the selection of projection nodes,the clustering of projection nodes as cluster heads,and the energy consumption of the projection data to the sink through routing nodes,which shortens the network lifetime.In addition,the measurement matrix is a key issue throughout the process of collecting compressed data.The existing measurement matrices have problems such as large randomness and difficulty in hardware implementation.Aiming at the above problems in the CDG with the selection of projection nodes,the establishment of joint routes,and the construction of the measurement matrix,the main researches in this paper are as follows:1.In view of the problem of random and uneven location of projection nodes in large-scale WSN,a compressive data gathering method based on even projection is proposed.(1)For the WSN with uniformly distributed of nodes,an even clustering method based on spatial location is proposed.The monitoring area is divided into grids with the same size.Each grid is examined according to the remaining energy of the nodes in the grid and the distance from the node to the sink node,and the optimal node is selected as the projection node(cluster head).Other nodes find the closest cluster head at the principle of the shortest distance,and forming a cluster.Compared with the random projection node method through the MATLAB simulation,this method ensures the equalization of the projection node's position,reduces the network energy consumption,and prolongs the network lifetime by about 25%.(2)For the WSN with uneven distribution of nodes,the location-based even clustering method consumes more energy on the grids with sparse node,thus an even clustering method based on node distribution density is proposed.Under the division of grids,the method adds the nodes in the grids whose density is lower than a certain threshold to adjacent clusters.Therefore,the grid ensures the balance of the projected node's position,and the node's density ensures the balance of the cluster,thus reducing the election of cluster heads and data transmission,equalizing energy consumption,and prolonging the network lifetime.Compared with the random projection node method for the network with the same scene through the simulation,this method has the obviously more number of remaining nodes in per round and the more excellent running status,and the network lifetime is extended by about 27%.Furthermore,the influence of parameters such as the number of nodes,the field's size,the compression ratio of the projection,the threshold of the node's density and etc.on the energy consumption of the network is studied.The MATLAB simulation results show that the node's density based even clusteringmethod is significantly better than the random projection node method and the random walk method.2.For the small-scale WSN,for using projection nodes generates additional energy consumption,a CDG method based on polar coordinates is proposed.This method uses polar coordinates to locate nodes,establishes chained structures to form routes,and applies random projection to collect compressed data,which is called Random projection-Polar coordinate-Chain routing(RPC)method.(1)Using virtual polar coordinates to locate the node.In the WSN,the sink is the center of data collection.With the sink as the pole,the orientation of each node relative to the sink can be clearly determined through polar coordinates,and it is easy to search for nodes according to certain conditions.(2)Adopting the chain topological structure.Starting from the node farthest from the sink,a chain is formed along the direction closer and closer to the sink.The node at the end of the chain acts as the cluster head,thus the end-node can transmit the weighted sum of the collected data to the sink by one-hop.And the chain route is established by a greedy algorithm,which significantly reduces the energy consumption and complexity than the tree structure.The RPC method abandons selecting projection nodes,and the cluster heads are naturally formed by the non-zero coefficient nodes,thereby reducing the energy consumption of electing projection nodes and establishing the tree routes,and making the entire network energy consumption balanced.(3)For smaller networks,a four-quadrant chain route method combining the polar radius and polar angles is used;For larger networks,a routing algorithm combining the sector and the inner-circle is used.According to the random projection theory,the weighted sum of the random projections of each row of the corresponding measurement matrix in each partition is transmitted to the sink.The sink has collected all measurements of each partition to complete the signal reconstruction.In this method,the route is formed by searching nodes in the zones according to the polar radius and the polar angle,which avoids the roundabout route between distant nodes and reduces the energy consumption of the network.By comparing the RPC method with other related methods(MSTP,GEM and PEGASIS)and the simulation experiments of different types of routes,it shows that the routing energy consumption based on the polar radius is 2 to 3 times than that of the four-quadrant route method,and the energy consumption based on the polar angle is about 1.2 times than that of the four-quadrant routes,and the energy consumption of this method is lower than that of the other three methods under different scales,indicating that both of the proposed methods can effectively reduce network energy consumption.3.Aiming at the problem that the hardware of measurement matrices is not easy to implement in the CDG for the WSN,a construction method of a dual-structure measurement matrix is proposed.Based on the theory of compressed sensing and sparse random projection,two different types of matrices are integrated into a new matrix,constructing a new type ofdual-structured measurement matrix of 'unit array + random square matrix'.Using this measurement matrix,the lower reconstruction error can be obtained than that of simply applying the random measurement matrix.Furthermore,a method of sub-framing and overlapping reconstruction is proposed to remove the large error caused by the unit array in the measurement matrix and ensure the stability of the entire signal reconstruction.This method is applied to the compression and reconstruction of the WSN signal and the speech signal.The simulation results show the SNR of the reconstructed signal using the proposed method is increased by about 10 dB and the reconstruction performance is significantly improved compared with the random measurement matrix.