Research On Compressive Sensing Based Data Aggregation In Wireless Sensor Networks

Wireless sensor networks consisting of a large number of small and low-cost sensor nodes are well-suited for various monitoring and measuring tasks in many applications.To accom-plish these targeted applications,sensor nodes have to collect and transmit a tremendous amount of real-time data over their lifetime.Due to the stringent energy constraints,weak computing ability and small storage capacity of sensor nodes,we need to establish an effi-cient data acquisition and transmission scheme to reduce the cost of information acquisition and prolong the lifetime of WSNs.As an economical data acquisition theory,compressed sensing provides a new data acquisi-tion approach for WSNs.Applying CS theory to traditional WSNs,sensor nodes can realize data acquisition in a compressed way with no requirement of additional computational over-head.In this way,the amount of data transmission could be greatly reduced,so does the consumption of network energy.Moreover,sensor nodes only afford the calculation of the compression part,which satisfies its characteristics of limited processing capability.And the reconstruction part with high computational complexity is carried out on the terminal computer,which is without the limitation of computing capacity and energy.However,due to the precondition of the compressed sensing theory,the data aggregation strategy based on the theory is different from the traditional method,and it is necessary to recover the original data from the compressed data.These two points are the key problems in the application of compressed sensing theory to WSNs.Considering the signals detected by a group of sensor nodes has spatial and temporal cor-relations,we construct a CS-based signal acquisition scheme which takes into fully account this characteristic to reduce the networks energy consumption,and thereby,to prolong the lifetime.In an arbitrary sampling cycler,each member node independently decides whether or not to collect and transmit its signal to FC,along with sampling cycle and node ID,with probability p_tx.In this way,it can be achieved for sensor nodes to sleep and work periodi-cally;consequently,network energy can be balanced well.We also propose a reweighted l₁-norm minimization algorithm via GPSR to reconstruct the original signal from noisy measurements,which are acquired by the former constructed CS-based signal acquisition scheme.The algorithm integrate the re-weighting procedure into each iteration,i.e.,the weights change according to the changes of solution in each iteration process.This method promotes the solution has the same sparsity structure which is present in the original signal.The simulation results show that our algorithm has a better performance in reconstruction accuracy and computation complexity.