Signal Sparsity Based Efficient Transmission Framework And Key Technology Research In IoT

The Internet of Things(IoT)is considered as the third industry development after the comput-er and the Internet.It has extensive applications,where information sensing and transmission play a vital role.Due to noise,environment change and interference,the communication between low power sensor nodes is unreliable.Conventional methods of communication use source coding to reduce the communication volume,then adopt the channel coding to resist data loss by increasing redundancy.However,existing source and channel coding induce lots of extra energy consump-tion and computation complexity.How to transmit sensing signal rapidly and accurately under unreliable communication and constraint resource is an urgent issue that needs to be tackled.Considering that there exist many sparse signals in the IoT,this dissertation applies compres-sive sensing(CS)theory to sparse signal transmission in the resource constraint IoT,aiming to promote the network performance.The main content includes:1.For spatial characteristics of lossy link,transitional region occupies a large area and com-munication in this region is unreliable.To fully utilize this region,a sparse signal transmission framework via lossy link using CS is proposed.Based on inherent sparsity of sensing signals,the data loss during transmission is modeled as a CS process.Fusion center can accurately reconstruc-t the original signal with received data through CS recovery algorithms.Furthermore,the packet length control problem in sparse signal transmission is exploited,where bursty data loss induced by larger packet length will deteriorate signal recovery performance.Data interleaving at sensor node is adopted for mitigating the bursty data loss effect.Experimental results show that the proposed method expands the reliable transmission region under lossy link transmission,and compared with traditional retransmission-interpolation methods,it can reduce the energy cost while promoting signal recovery performance.2.To deal with the data collision when several devices access the channel,the structured sparse signal random access is proposed to jointly recover several original signals that share a common support.Measurement matrix construction is implemented by reducing the multiple mea-surement vector(MMV)problem to a single measurement vector(SMV)problem,where packet loss during random access is modeled as a CS process and measurement matrix can be constructed based on correctly received data.The sensing probability is further employed to control transmit-ting data number and randomize data loss.The optimal sensing probability is derived to minimize energy consumption by joint consideration of random access transmission and signal recovery requirements.Simulation results validate the effectiveness of SSSRA in utilizing the structured sparsity and show its improvement compared with other CS based methods.3.For existing compressive data gathering(CDG)methods,data transmission number is still large to produce single measurement number,and the robustness of CDG with node failure still need to be improved.To solve these problems,a cluster based sparsest compressive data gathering(CS-CDG)method is proposed.Compared with existing CDG method,data senses by sensor node can be considered as measurement data based on the partial basis,reducing data transmissions in clusters.Note that son nodes can communicate with CH by relaying through other son nodes,or directly transmitting by power control.The analytical models that study the relationship between the size of clusters and the energy cost using these two intro transmission methods is proposed,aiming at finding which method is better.Also,the optimal size of clusters that can lead to mini-mum energy cost is derived under the analytical models.When node failure happens,our proposed CS-CDG can still keep the system performance by replacing the broken nodes since measurement data is only correlated with one data.Extensive simulations confirm that our method can reduce the number of transmissions significantly and the robustness of our system is much better.4.For CS applications,dense projection matrix such as Gaussian random matrix and Bernoul-li random matrix will induce lots of sensing and transmission cost,while the sparsest projection matrix has the limit that the nonzero element in the matrix must be random to meet the RIP con-straint.Through theoretical analysis,it is proved that with only two or three numbers conforming to Gaussian distribution in each row of the projection matrix,the RIP can be satisfied.Compared with traditional random Gaussian or Bernoulli matrices,the sparse Gaussian matrix can guarantee recovery performance with less memory.Further the sparse signal transmission via lossy link is used as an example to show its superiority.