Research Of Data Collection Technology Based On Compressive Sensing In Wireless Sensor Network

As one of important components of the Internet of Things, wireless sensor network has been widespread concern in academic circles. The wireless sensor network consists of a large number of micro-nodes with limited energy, so one of key issues for wireless sensor network research is how to reduce the energy consumption. The proposal of the theory of compressed sensing replace the traditional Nyquist sampling theorem. Its characteristics of simple encoding side and complex decoding side provide a new way to solve the problem of data collection in wireless sensor networks. In this dissertation, we mainly did the following works:(1) Firstly, we started from the basic principles of compressed sensing, summed up the requirements for the measurement matrix structure, analyzed the common deterministic matrix: Toeplitz and Circulation. Then we proposed a novel random spacing sparse matrix generated by a seed vector which can be got from the diagonal matrix created by applying SVD in Bernoulli matrix. Compared to Toeplitz matrix, the new matrix has less independent variables and more focused matrix information. The experiments also show that it has better effect of reconstruction. We also apply it in data collection of chain wireless sensor network, and the result of experiments show that it has a significant reduction in packet forwarding capacity of network, thus reduce the network energy consumption and finally prolong the network life cycle compared to traditional data collection in wireless network.(2) Due to the limitation of communication distance of nodes in network, the node may not be able to reach the sink node by single hop, a novel data collection method based on random spacing sparse compressed sensing proposed in the cluster wireless sensor network. We split the cluster wireless sensor networks into two parts: the part within the cluster, the cluster head collects data generated by using compressed sensing in which matrix is generated by the seed vector sent from the sink node. The part between clusters, cluster head forward measurement values to the sink node by using the former constructed multi-hop routing. Finally, the result of comparison and analysis of its performance shows that the new method has less energy consumption in a cluster or inter-cluster.