面向稀疏信号的物联网高效传输体系及关键技术研究

发布时间：2018-03-26 12:13

本文选题：物联网稀疏信号　切入点：压缩感知　出处：《浙江大学》2017年博士论文

【摘要】：物联网被视为继计算机、互联网之后世界信息产业发展的第三次浪潮,应用前景巨大。其中信息感知与传输在物联网应用中起着至关重要的作用。但由于噪音及干扰等因素的影响,物联网中低功耗节点之间的数据传输并不可靠,而传统的信源、信道编码通常为节点端引入额外的能量开销及复杂的计算。如何在不可靠传输及资源受限的情况下,快速准确地传输原始信息是一个亟待解决的问题。考虑到物联网中众多的原始信息具有稀疏性,本文将压缩感知理论应用于资源受限物联网的稀疏信号传输体系中以提高网络性能。其主要研究内容包括以下部分:1.针对有损链路过渡区范围广而数据传输并不可靠的问题,本文利用信号固有的稀疏特性,将有损传输过程中的数据丢失模拟为随机压缩采样。接收端通过接收到的部分数据即可通过重构算法恢复原始稀疏信号。进一步考虑到长数据包传输导致的块状丢包不利于信号重构,我们在节点端进行交织以随机化数据丢包,保证了信号的传输性能。实验结果表明,该方法大大拓宽了有损链路空间可用范围,且相对于传统的数据丢失重传-插值方法能够有效减少能耗并提高重构精度。2.针对若干传感器节点信道接入问题,本文利用感知信息的结构化稀疏特性,提出了结构化稀疏信号随机信道接入。考虑到传统多量测向量模型中投影矩阵构造不适用于该场景下,我们将多量测向量问题转化为单量测向量问题。并在节点端引入感知概率的概念以控制通信量,通过联合考虑随机信道接入及信号重构需求,求得最优感知概率以最小化系统传输能耗。3.针对现有的压缩数据收集耗能较大的问题,本文提出了基于分簇结构的最稀疏压缩数据收集。相对于现有的数据收集方法,该机制中每个量测值只需一个节点的数据,大大减少了传输量。注意到数据传输可通过调节功率直接传输或中间节点转发两种方式,本文分别对其能耗建模以进行比较。并在上述能耗模型下获取最优簇大小。该方法能够以较少的能耗进行压缩感知下的数据收集,且节点故障时鲁棒性较好。4.针对压缩感知应用中稠密投影矩阵导致高感知及传输能耗的问题,本文提出一种简单易构造的稀疏高斯矩阵。通过理论分析,我们证明了该矩阵每行只需少数高斯随机数即可满足约束等距性,保证重构算法高精度恢复原始稀疏信号。相对于传统的随机矩阵,该矩阵能够在节点端消耗较少时间及内存的同时保证压缩重构性能。进一步以有损链路下稀疏信号传输为例说明该矩阵的优势。
[Abstract]:The Internet of things is regarded as the third wave of the world's information industry after computers and the Internet. The application prospect is huge. The information perception and transmission play an important role in the Internet of things application. However, due to the influence of noise and interference, the data transmission between the low-power nodes in the Internet of things is not reliable, and the traditional information source. Channel coding usually introduces additional energy overhead and complex calculations for the node. Fast and accurate transmission of original information is an urgent problem. Considering the sparsity of many original information in the Internet of things, In this paper, the theory of compressed sensing is applied to the sparse signal transmission system of resource-constrained Internet of things to improve the network performance. The main research contents include the following parts: 1. Aiming at the problem that the lossy link transition area is wide and the data transmission is unreliable, In this paper, we use the inherent sparse characteristic of the signal. The data loss during lossy transmission is simulated as random compression sampling. The receiver can recover the original sparse signal by reconstructing some of the received data. Further consideration is given to the block caused by long packet transmission. Packet loss is not conducive to signal reconstruction, In order to ensure the transmission performance of the signal, we interleaved at the node end to randomize the data packet loss. The experimental results show that the proposed method greatly broadens the available range of lossy link space. Compared with the traditional data loss retransmission interpolation method, it can effectively reduce the energy consumption and improve the reconstruction accuracy. 2. Aiming at the channel access problem of some sensor nodes, this paper uses the structured sparse characteristic of perceptual information. The structured sparse signal random channel access is proposed. Considering that the projection matrix construction in the traditional multi-measurement vector model is not suitable for this scenario, We transform the multi-measurement vector problem into the single-measure vector problem, and introduce the concept of perceptual probability to control the traffic at the node end, and consider the requirements of random channel access and signal reconfiguration. The optimal perceptual probability is obtained to minimize the transmission energy consumption. 3. Aiming at the problem that the current compressed data collection consumes a lot of energy, this paper proposes the most sparse compressed data collection based on clustering structure. Compared with the existing data collection methods, In this mechanism, only one node is required for each measurement, which greatly reduces the amount of transmission. In this paper, the energy consumption is modeled for comparison, and the optimal cluster size is obtained under the above energy consumption model. This method can collect data under compressed sensing with less energy consumption. In view of the problem that dense projection matrix leads to high sensing and transmission energy consumption in compressed sensing applications, a simple and easy to construct sparse Gao Si matrix is proposed. It is proved that this matrix needs only a few Gao Si random numbers per row to satisfy the constraint equidistance, which ensures that the reconstruction algorithm can restore the original sparse signal with high accuracy. The matrix can consume less time and memory at the node end and ensure the compression and reconstruction performance. Further, the sparse signal transmission in the lossy link is taken as an example to illustrate the advantage of the matrix.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP391.44;TN929.5

【相似文献】