Study On Compressed Sensing Technology Of Wireless Sensor Networks

Wireless sensor networks (WSNs), as an important part of the Internet of things, has been widely used in many related fields, such as environmental monitoring, medical testing, and military deployment. However, there are also many resource constraints of sensors, which bring a lot of restrictions for large-scale deployment and long time running of WSNs. As a new signal sampling theory, compressed sensing theory not only breaks the limitations of traditional Shannon sampling theory in terms of the sampling frequency of sparse signal or compressible signals, but also has the advantages of compression and sampling can be performed simultaneously. In WSNs, adopting compressed sensing theory to collect data is mainly using the spatial correlation that exists among the sensing data, using the measured matrix to make the sensing data become sparse, allowing the network to transmit data packets from N reduced to M (M << N), which can reduce the entire network communication cost without producing a lot of computation or transmission overhead, and also ensure the accuracy of data recovery, prolong the lifetime of the network. The existing compressed data collection projections in WSNs, usually randomly generate the measurement matrix by the Sink, and then generate the data collecting route based on the measurement matrix passively. Because the routing is not according to the network structure, therefore there is little help to improve network performance, and some nodes need to use multiple relay nodes to achieve data compression, which not only increases the energy consumption of nodes, but also makes the node energy consumption is not balanced.To solve the above problems, with the goal to maximize the reduction and balance the network energy consumption, this paper propose a compressed data collecting scheme based on the depth-first spanning tree in WSNs. The detail research contents and results are as follows:(1) This paper describes the characteristics of WSNs and data collecting problems in it. The WSNs has the advantages of small size, low cost, flexible deployment, etc., but at the same time it is a resource-constrained network, which brings many difficulties in WSNs data collection. Although traditional data integration projections can reduce the energy consumption of nodes, prolong network’s running time, but still can’t meet people’s demands for WSNs.(2) This paper reviews the basic theoretical framework of compressed sensing, summarizes the problems in existing CS-based WSNs compressed data collection projections. According to the compression sensing theory, if the signal is sparse or compressible, it can be accurately restored from small amount of compressed data which contains all information. It breaks through the limitations on sampling frequency from traditional Shannon sampling theory, reduces the sampling frequency of the signal, and simplifies the process of data compression. However, the existing research on methods of data collection based on compressed sensing theory in WSNs applies the theory simple. And how to well apply the compressive sensing theory to the WSNs to collect data is also facing many challenges.(3) The proposed projection in this paper collects data according a depth-first spanning tree. It generates the routing tree according to the characteristics of the network deployment, and then builds the measurement matrix according to the routing tree The specific method is to generate M (group number) sub-trees from Sink firstly, and the nodes in the same sub-tree connected directly. every sub-trees as one group, the nodes compress and transfer data within the same group, since each node transmits data only once, thereby achieving a load balancing within the nodes in same group. The compressed data finally transfers from the root of the sub-tree to the Sink, because the data transfer process fully considers many factors, such as the transmission number and the shortest path, thus reducing the number of relay nodes, minimizing the energy consumption introduced by data transfer between groups, while the introduced energy consumption dispersed on the relay node evenly, balancing the node energy consumption. In addition, data collecting tree of the projection has better network performance because it is generated based on the characteristics of the network deployment.(4) The simulation results show the effectiveness of the proposed scheme. Using self-designed simulation platform, this paper compares the DFST projection with the minimum transmission data collection tree MTT projection and cluster-based data collection CCS projection in overall energy consumption and node load balancing. Simulation results show that the proposed scheme has better energy efficiency and load balancing features.