Research On Data Gathering Based On Compressive Sensing In Wireless Sensor Networks

Wireless Sensor Networks(WSNs)have been widely used to gather data periodically in the monitoring environment through cooperation between sensor nodes.However,battery-powered nodes are limited in energy and data processing capability is weak.Additionally,the densely deployed nodes cause a large amount of redundant information in the network,the node deployment environment is complex and diverse,the wireless links is unreliable,and the packet loss rate is high.These characteristics limit the development of WSNs.The application of CS in WSN solves the problem of the energy hole,reduces the transmission energy of nodes and prolongs the network life.Nevertheless,most of the current CS-based data gathering studies aim to reduce energy consumption and do not consider the problems in the actual network environment,which such as the deployment environment has a high packet loss rate,the reconstruction accuracy is greatly affected by packet loss,and the continuous cycle gathering data has high time-space redundancy.Therefore,simply assume that the node is deployed in an ideal environment,the wireless link is completely reliable,and a fixed route can be established with strong robustness,which does not correspond to the actual situation.The research of this dissertation relies on the national major special project to reduce the energy consumption of nodes,prolong the life of the network,and reduce the impact of packet loss rate on the CS reconstruction accuracy.Moreover,this dissertation has conducted an in-depth study on the theoretical application of CS-based WSNs data gathering methods in different practical application scenarios.The main contributions of this thesis are summarized as follows:1.In cluster routing topology and the scenario of the single cluster and single data gathering period,aiming at the problem that the performance of existing CS-based data gathering algorithms is sensitive to packet loss,this dissertation deeply studies the impact of packet loss on the data reconstruction accuracy.By analyzing the real data matrix,a sparse block measurement matrix is proposed,which makes it possible to apply the Matrix Completion(MC)theory to recover the missing measurements and reduce the energy consumption of each measurement gathered by cluster head.Since this,combined with MC and CS theory,the CS data gathering algorithm based on sparse block measurement matrix is proposed.Sink first recovers the missing data based on MC theory to guarantee the integrity of the observation matrix and reduce the impact of packet loss on CS measurements.Secondly,the CS theory and the sparse block measurement matrix are used to reconstruct the data,which reduces the number of measurements and prolongs the network lifetime.By modeling the energy consumption of sending and receiving data packets,the optimal number of clusters in the proposed algorithm is given.The theoretical analysis and simulation results show that the proposed algorithm effectively improves the reconstruction accuracy when the packet loss rate is less than 50%.2.In the scenario of the multiple clusters and single data gathering period,the dissertation aims at the problem that the measurement matrix can not be updated according to the packet loss in real-time,which causes the joint reconstruction accuracy of the whole network data to be affected by the packet loss rate.Therefore,a data-gathering algorithm is proposed to reduce and balance the network energy consumption and reduce the impact of packet loss.The member nodes randomly send the gathered data to the cluster head with a probability.Then the cluster head reversely formulates a random sparse measurement submatrix according to the received data and updates this matrix in real-time after each round of data gathering to avoid measuring the lost node.To employ spatial correlation between clusters,the sink constructs a block diagonal matrix as a measurement matrix via the submatrices and uses it to reconstruct the entire network data.Additionally,the expression of total energy consumption is given and the optimal number of clusters is discussed under this framework to reach the minimum power consumption.Finally,the proposed algorithm is evaluated on the emulated data and the real sensor data,respectively.The theoretical analysis and simulation results indicate that the proposed algorithm reaches high precision and increases the number of data gathering rounds under reliable links.Moreover,the impact of packet loss on data reconstruction accuracy is effectively reduced.3.In the scenario of the multiple clusters and multiple data gathering periods,the dissertation aims at the problem that how to exploit spatial and temporal correlations of data simultaneously to reduce the number of measurements,and how to reduce the impact of packet loss on the reconstruction accuracy.Therefore,this dissertation proposes an algorithm that combines Kronecker Compressed Sensing(KCS)and cluster topology to utilize spatiotemporal correlation and reduce the impact of packet loss.Temporal and spatial measurement matrices are proposed by modeling data in time and space,respectively.Furthermore,the Kronecker product of the two matrices is used as the space-time measurement submatrix by cluster head.Then,the sink constructs those matrices as a block diagonal matrix to utilize the spatial correlation among clusters.Moreover,the energy consumption model is established and the network energy consumption is analyzed to give the optimal cluster number and data sampling rate under the proposed framework.The theoretical analysis and simulation results present that compared with the existing algorithms,the proposed algorithm can effectively balance the performance-energy trade-off in both lossy link and ideal link.4.In the scenario where the node in the network does not establish a fixed route and the sink is mobile,this dissertation proposes a data gathering algorithm,which exploits the distributed data storage and the broadcast properties of the wireless channel,to transmit the broadcast data packet orderly and finally converges.Partial nodes are randomly selected as source nodes to broadcast their own packets.Then,the neighbor nodes of those source nodes receive and merge those broadcast packets under certain conditions,and broadcast the merged data packets with a probability in order to reduce the number of transmissions and receptions.After the broadcast process is completed,the sink randomly visits some nodes and extract the corresponding data.Then the sink constructs a sparse measurement matrix and the measurement vector according to the data,and reconstructs the entire network data.In this way,the impact of packet loss is reduced.Additionally,the property that the rows of the measurement matrix are linearly independent is demonstrated.To analyze the energy efficiency of the proposed algorithm,the expression of the total number of transmission and receptions is derived based on random geometric graph theory.The theoretical analysis and simulation results show that the proposed algorithm has high reconstruction precision,effectively reduces the number of data packet transmission,reception and merging,and fast convergence in the data broadcasting phase,effectively reduces the impact of packet loss.In conclusion,this dissertation aims at optimizing network energy consumption and improving data reconstruction accuracy under the unreliable link.This dissertation researches on data-gathering algorithms for fixed-route scenarios with a single cluster and a single period,multiple clusters and a single period,multiple clusters and multiple periods,and non-fixed-route scenarios with random broadcasting,respectively.These studies range from simple scenario to complex scenario,from spatial correlation to temporal correlation,from static routing to dynamic routing.