Compressive Data Gathering Algorithms In Wireless Sensor Networks

Wireless Sensor Networks(WSNs)have the features of easy deployment,being distributed,self-organization and etc.It has a wide application prospect in the fields of the environment monitoring.Because energy of sensor nodes is limited,the effective data gathering in the network monitoring area has become the most urgent problem to be solved.Compressive Sensing(CS)is a novel signal sampling theory.It first gathers fewer measurements by the measurement matrix,and utilizes the reconstruction algorithm to recover original data.The method can reduce the computational complexity of sensor node coding,and balance the energy of sensor nodes in the network.Therefore,CS theory provides an effective gathering framework for sparse data in WSNs or data which are sparse in a transformation domain.In the abnormal temperature detection and temperature monitoring in WSNs,the abnormal temperature data are sparse relative to a large series of temperature data,and the temperature data are usually sparse in a certain transform domain.Based on the research background,the dissertation further studies sparse data gathering and sparse representation of data.The main contents are as follows:Sparse data gathering algorithm in WSNs is studied to realize abnormal temperature detection in the environment monitoring.Because the temperature in the monitoring area changes slowly,abnormal data caused by sudden changes in temperature can be considered as a sparse event.For the problem of sparse event detection,the dissertation considers the mobility and intelligence of mobile agent,and proposes a mobile agent based sparse data gathering algorithm(MA-Greedy).The algorithm adopts sparse binary matrix with a fixed number of nonzero entries in each column as the measurement matrix,which guarantees the recovery performance of sparse data,and introduces mobile agent model into sparse data gathering under the CS framework.According to Coefficient of Variation,MA-Greedy algorithm plans the migration paths of mobile agents by the measurement matrix and greedy strategies,and improves the balance performance of energy consumption among sensor nodes.Adaptive sparse data gathering algorithm is studied to realize the continuous detection of abnormal temperature in the environment monitoring.In practical applications,the sparsity of sparse data is unknown.For the problem,the dissertation studies the estimation method of sparsity based on sparse binary matrix,and further studies the adjustment problems of the minimum number of measurements on account of sparsity and the length of data.Finally the dissertation puts forward a sparsity estimation based sparse data gathering algorithm.In each round of data gathering,the algorithm first adopts the sparsity estimation mechanism to accurately estimate the sparsity of sparse data,and uses Monte Carlo experiments to determine the function relationship between the minimum number of measurements and the sparsity in order to adaptively adjust the number of measurements under the circumstance of guaranteeing the performance of data recovery.At the same time,the dissertation adopts total energy consumption of the network and the balance of energy consumption between sensor nodes as synthetical evaluation index,and proposes an information entropy based synthetical evaluation method.Simulation experiments validate synthetical performance of proposed algorithm.A sparse representation algorithm for sensory data in temperature field is studied to realize the application of CS theory in temperature field monitoring.For the sparsity problem of sensory data in the temperature field,the dissertation presents a diffusion wavelet based sparse representation algorithm for sensory data in WSNs.The algorithm describes the spatial correlation among sensory data through weighted graphs,and applies diffusion wavelet theory for multi-scale analysis of sensory data.Diffusion operator is constructed based on the communication radius,weighted adjacency matrix and normalized Laplace matrix.It has been proved that high powers of the constructed diffusion operator have low numerical rank in theory,and they can be compressed in the subspace.The synthetic temperature field experiment and real temperature field experiment verify that constructed sparsifying basis can make sensory data sparser.Compared with other sparsifying bases,the constructed sparsifying basis decreases reconstruction error of sensory data.Sparse representation and gathering algorithm of sequential temperature data in environmental monitoring are studied to realize the continuous monitoring of environmental temperature.For the gathering problem of sequential temperature data,the dissertation proposes a compressive measurement model based on the spatial correlation between sequence temperature data.Through the model,the sparsifying basis of sensory data and the measurement matrix are built.The dissertation puts forward a sequential data gathering algorithm based on CS to minimize the total energy consumption of the network.Based on this assumption the temperature data has short-term stability,the model uses eigenvalue decomposition of the covariance matrix to construct the sparsifying basis of sequential data.The algorithm adopts sparse binary matrix as the measurement matrix and designs the shortest path algorithm.In each measurement,only part of sensor nodes participate in compressive data gathering,and their sensory data are sent to sink node by the shortest path algorithm.The environmental temperature data confirm the effectiveness of compressive measurement model,and the shortest path algorithm reduces total energy consumption of the network.