Research On Data Collection Strategy Of Large Scale Wireless Sensor Networks Based On Bayesian Compressed Sensing

Internet of things(IOT)is known as the 4.0 industrial revolution and wireless sensor networks(WSNs)are the most important part of IOT.WSNs are widely used in military and civil fields by collecting data from a large number of sensor nodes randomly deployed in the monitoring area and cooperating with each other in the form of ad hoc network.But the energy of sensor nodes is limited.Once the energy is exhausted,it will lead to the end of the life of WSNs.Compressed sensing(CS)creates a new way of WSNs data collection.Compared with the traditional signal compression processing method,there is no need to compress the original signal on the sensor node.Instead,a handful of measurement signals are generated on the sink node by random projection and the original signal is perfectly reconstructed by the designed reconstruction algorithm.Therefore,the application of CS technology in large-scale WSNs data collection scene can break through Nyquist sampling rate.It can reduce the energy consumption of WSNs and extend the service life of WSNs.But at present,the WSNs data collection based on CS does not consider the actual application scenarios of WSNs too much,such as: unreliable wireless communication link,data packet loss,noise interference,node energy consumption,and so on.Therefore,the research of large scale WSNs data collection strategy based on Bayesian CS in this paper consider the problems of WSNs actual application scenarios.In order to improve the reconfiguration performance of CS and extend the service life of WSNs,the paper study intensively the adaptive sparse basis and active node selection.The main innovations are as follows:(1)Aiming at the problems of unreliable communication link,noise interference and changeable environment of monitoring area in WSNs,in order to ensure the effective data collection of WSNs based on CS and prolong the service life of WSNs,the paper propose an adaptive sparse representation of the original signals sensed by WSNs and an adaptive sparse strategy based on Bayesian CS.This strategy can not only improve the reconfiguration performance of CS and ensure the effective data collection of WSNs,but also further reduce the energy consumption of WSNs and prolong the life cycle of WSNs.Firstly,the measurement matrix is constructed by combining sparse routing and Bernoulli random matrix.The noise interference is considered and the compressed sensing model is updated.The sparse Bayesian learning method is adopted to perfectly reconstruct.Then,the optimal updating of sparse basis is obtained by minimizing the posterior covariance reconstructed by Bayesian method.In order to satisfy the RIP condition for the combination of the optimized sparse basis and the measurement matrix,the paper adopt the low coherence principle to replace the RIP condition.Finally,through the analysis and reasoning of the problem,the paper obtain the optimized update representation of sparse basis,and design the corresponding algorithm to get the final sparse basis representation.The simulation results show that this strategy has better reconstruction performance than DCT,wavelet transform and K-SVD algorithm.Compared with dense routing,it can further reduce the energy consumption of WSNs and prolong the life cycle of WSNs.(2)For the application scenario of large-scale WSNs data acquisition,the original signal sensed by WSNs has huge redundancy.Therefore,most of the sensor nodes in WSNs can be set into sleep state,and only a handful of active sensor nodes could still complete the data acquisition task of WSNs.In order to ensure the data acquisition quality of WSNs and further reduce the energy consumption of WSNs,the heterogeneous sensing environment and different quality of sensing data of each sensor node are considered.By optimizing the active sensor nodes in WSNs,the paper propose an active node selection strategy based on Bayesian compressed sensing,which could withstand the impact of certain network data packet loss and reduce the energy consumption of WSNs as much as possible under the condition of ensuring the reliability of network data collection.Firstly,according to the characteristics of heterogeneous sensor environment,the sampled data on sink node is interfered by the noise.The paper update the compressed sensing model and recover perfectly through sparse Bayesian learning.Then,the paper design the active node selection framework by minimizing the posterior covariance reconstructed by Bayesian method.Finally,the active node selection framework is optimized,with considering the energy constraints of each sensor node.The simulation results show that the strategy has better reconfiguration performance than BP,Jeffrey,OMP and other algorithms.Compared with the distributed random node selection scheme,it can further improve the life cycle of WSNs under the same reconfiguration performance.