基于压缩感知的无线传感器网络信息处理与传输机制研究

发布时间：2018-05-03 21:02

本文选题：无线传感器网络 + 压缩感知　；参考：《上海交通大学》2014年博士论文

【摘要】：无线传感器网络是以数据为中心的网络,其主要目的是从监测区域内收集感知对象的信息,并对其进行处理,以尽可能少的能耗传输到数据管理中心。然而,无线传感器网络节点数量众多,分布密集,节点资源(包括能量、通信、计算和存储能力)受限,如何对感知数据进行处理及高效传输是无线传感器网络中的核心问题。针对无线传感器网络自身独特的特点及传统传感信息处理与传输方法的不足,本文提出了基于压缩感知的传感信息处理与传输机制,不仅简化了节点信息处理的复杂度,降低了对计算资源的要求,而且克服了传统压缩算法中信息处理的不对称性。本文根据无线传感器网络中的不同应用需求,以提高网络容量、降低数据传输时延以及减少整个网络的传输能耗为设计目标,着重从传感数据采样方式、网络路由协议设计、节点调度策略设计及网络性能分析等几个方面进行深入研究,探讨一种无线传感器网络的信息处理与传输的新机制。本文的主要研究内容概括如下:1.基于压缩感知的大规模无线传感器网络的数据收集本文首先将压缩感知理论引入大规模无线传感器网络的数据收集应用中,研究了单汇聚节点和多汇聚节点的数据收集网络的网络容量及传输时延问题。针对单汇聚节点的数据收集网络,给出了基于压缩感知框架下数据收集网络的网络容量上界,提出了一种最优的网络容量下界的路由与调度策略,分析了数据传输时延性能。针对多汇聚节点的数据收集网络,首次引入了稀疏随机投影理论,提出了基于压缩感知的多会话数据收集方法;给出了多汇聚节点的数据收集网络的网络容量上界,构造了一种多会话生成树并提出了最优的网络容量下界的路由与调度策略,分析了数据传输时延性能。理论分析及仿真结果表明,压缩感知方法能大大提高大规模无线传感器数据收集网络的网络容量及降低数据传输时延。2.基于压缩感知的的网间计算基于压缩感知的数据传输在传输过程中通过将节点间的数据转发转化为节点间的数据计算,从而降低整个网络的传输能耗。因此,基于压缩感知的数据收集方法对降低无线传感器网络的传输能耗到底带来多大的优势是一个值得研究的课题。本文将压缩感知理论中对随机投影的构造转化为对一个多轮随机线性目标函数的计算,提出了基于树结构以及基于流言的计算协议来实现基于压缩感知的网间计算,首次从网间计算的角度评估压缩感知在传输能耗及传输时延上的性能表现。针对基于树结构的计算协议,提出了在最优的计算更新速率下的路由及调度策略,以及考虑传感数据的时间相关性时用于进一步提高计算性能的块计算协议。针对基于流言的计算协议,提出了一种广播流言算法,以实现网络拓扑易变情况下的信息传输。理论分析及仿真结果表明,本文提出的基于压缩感知的计算协议能有效减少网络的传输能耗及降低数据传输时延。3.基于压缩感知理论及随机游走的数据收集本文在压缩感知理论基础上,提出了一种基于随机游走的无线传感器网络数据收集算法。首次从图论、马尔可夫链理论及压缩感知理论等理论角度研究了该算法可行性的理论依据,给出了随机游走路径步长及所需的随机游走路径数等重要参数,并分析了基于?1范数最小化算法进行信号重构的理论依据。该算法突破了传统压缩感知理论中节点需均匀采样的限制,为压缩感知理论在无线传感器网络中的应用提供了一种更切实可行的方法;与基于传统压缩感知理论的收集方法相比,具有占用存储空间小、计算复杂度低以及传输能耗低等优点。
[Abstract]:Wireless sensor networks (WSN) is a data centric network. The main purpose of the network is to collect and process the information of the perceived objects from the monitoring area, and to transmit it to the data management center with as little energy as possible. However, the nodes of the wireless sensor network are large and dense, and the node resources (including energy, communication, computing and storage) are dense. The core problem of wireless sensor networks is how to handle and transmit the perceptual data efficiently. In view of the unique characteristics of the wireless sensor network and the shortage of traditional sensing information processing and transmission methods, this paper proposes a sensing information processing and transmission mechanism based on compressed sensing, which not only simplifies the node letter. The complexity of interest processing reduces the demand for computing resources and overcomes the asymmetry of information processing in traditional compression algorithms. According to the different application requirements in wireless sensor networks, this paper aims to improve the network capacity, reduce the delay of data transmission and reduce the transmission energy consumption of the entire network. According to the methods of sampling, network routing protocol design, node scheduling strategy design and network performance analysis, a new mechanism of information processing and transmission of wireless sensor networks is discussed. The main contents of this paper are summarized as follows: 1. data collection of large-scale wireless sensor networks based on compressed sensing In this paper, the compression perception theory is introduced into the data collection application of large-scale wireless sensor networks. The network capacity and transmission delay of the data collection network with single aggregation nodes and multiple converging nodes are studied. The network of data collection network based on the compressed sensing framework is given for the data collection network of single aggregation nodes. An optimal routing and scheduling strategy for the lower bound of network capacity is proposed. The performance of data transmission delay is analyzed. The sparse random projection theory is introduced for the first time in the data collection network of multi aggregation nodes, and a multi session data collection method based on compressed sensing is proposed, and the data collection of multiple aggregation nodes is given. In the upper bound of network capacity, a multi session generation tree is constructed and the optimal routing and scheduling strategy of the network capacity lower bound is proposed. The performance of data transmission delay is analyzed. The theoretical analysis and simulation results show that the compressed sensing method can greatly improve the network capacity and reduce the data of the large scale wireless sensor data collection network. Transmission delay.2. based on compressed sensing based inter network computing, data transmission based on compressed sensing is converted into data computing by transferring data between nodes in the transmission process, thus reducing the energy consumption of the entire network. Therefore, the data collection method based on compressed sensing is used to reduce the transmission energy of Wireless Sensor Networks. In this paper, this paper transforms the construction of random projection into a multi wheel random linear objective function in the compression perception theory, and proposes a algorithm based on tree structure and a rumor based computing protocol to compute the Internet based on compressed sensing. The performance of compressed sensing on transmission energy consumption and transmission delay is evaluated. The routing and scheduling strategy at the optimal computing update rate are proposed for computing protocol based on tree structure, as well as the block computing protocol, which is used to further improve the computing performance when the temporal correlation of sensing data is considered. A broadcast gossip algorithm is proposed to achieve information transmission in a network topology. The theoretical analysis and simulation results show that the proposed compression based computing protocol can effectively reduce network transmission energy consumption and reduce data transmission delay.3. based on compressed sensing theory and random walk data collection. In this paper, based on the theory of compressed sensing, this paper presents a data collection algorithm for wireless sensor networks based on random walk. The theoretical basis of the feasibility of this algorithm is studied from the theory of graph theory, Markov chain theory and compression perception theory. The steps of random walking path and the number of random walk paths are given. The theoretical basis of signal reconstruction based on the 1 norm minimization algorithm is analyzed. The algorithm breaks through the restriction of uniform sampling in the traditional compressed sensing theory, and provides a more practical method for the application of compressed sensing theory to wireless sensor networks; and based on the traditional compression perception theory. Compared with the collection method, it has the advantages of small storage space, low computational complexity and low transmission energy consumption.

【学位授予单位】：上海交通大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP212.9;TN929.5

【共引文献】