无线传感器网络中基于网络结构的分布式估计研究

发布时间：2018-07-10 05:03

本文选题：分布式估计 + 无线传感器网络　；参考：《西南大学》2017年硕士论文

【摘要】：分布式估计的目的是给定一个观测序列,网络中的节点通过合作的方式来估计一个随机或者确定性的参数。由于分布式估计算法的稳定性、鲁棒性和节能性等特点,使得其在无线传感器网络中非常实用。分布式估计主要有三种信息交换策略,扩散策略、一致性策略及增量策略,其中扩散策略的估计性能最优。无线传感器网络(WSN)在不同的空间地点采集观测数据,可以获得更大的平均信噪比,通过分布式处理大量的采集信息能够提高估计的精确度,提高鲁棒性,而且其拓扑结构的独特的特点对于许多应用有非常重要的意义。无线传感器网络的结构是分布式估计的基础,分布式估计算法是分布式估计的核心,因此,将网络结构与分布式估计算法有效的结合起来,会更有效的解决分布式参数估计问题。扩散最小均方算法(DLMS)是典型的分布式估计方法,由于DLMS具备结构简单、易于实现、性能稳定、鲁棒性强等特点,使得DLMS算法的应用较为广泛。然而DLMS算法也存在缺点,网络中的节点都要接收和发送数据直接给与自己相连的邻居节点,那么节点间总的通信量会有负担。本文首先探究了表征WSN网络局部结构的特征量——模体(包括三节点模体和四节点模体)对DLMS算法性能的影响,发现DLMS算法的性能与网络中带有闭合三角行模体数量有一定关系。进而针对DLMS算法节点间通信负担重的问题,提出了打破模体扩散最小均方算法,此算法大大减少了节点间的通信负载,且算法估计性能损失较小,更好的达到了通信负载与估计性能的均衡,这个研究对于节约网络能量和带宽有重要作用。本文还首次将扩散策略应用到相位估计中,结合交替迭代最小化方法,提出了基于传感器网络的分布式相位估计算法,提出的算法能更好的抗击噪声的干扰。接着本文从网络的整体结构出发,探究了WSN中不同的网络模型,包括规则网络、小世界网路、随机网络和无标度网络,对提出的分布式相位估计算法性能的影响。发现采用不同的网络模型,得到的算法性能有较大差异,这个研究对于分布式参数估计问题中传感器网络的拓扑结构设计有一定指导作用。
[Abstract]:The purpose of distributed estimation is to estimate a random or deterministic parameter by means of cooperation between nodes in the network given an observation sequence. Because of the stability, robustness and energy saving of the distributed estimation algorithm, it is very practical in wireless sensor networks. There are three kinds of information exchange strategy, diffusion strategy, consistency strategy and incremental strategy, among which the estimation performance of diffusion strategy is optimal. Wireless sensor networks (WSN) can obtain greater average SNR by collecting observation data at different spatial locations. The estimation accuracy and robustness can be improved by distributed processing of a large amount of collected information. Moreover, the unique characteristics of its topology are of great significance to many applications. The structure of wireless sensor networks is the basis of distributed estimation, and the distributed estimation algorithm is the core of distributed estimation. It can solve the problem of distributed parameter estimation more effectively. Diffusion least mean square algorithm (DLMS) is a typical distributed estimation method. DLMS is widely used because of its simple structure, easy implementation, stable performance and strong robustness. However the DLMS algorithm also has its shortcomings. The nodes in the network have to receive and send data directly to their neighbor nodes so the total traffic between the nodes will have a burden. In this paper, the influence of characteristic motifs (including three-node motifs and four-node motifs) that characterize the local structure of WSN networks on the performance of DLMS algorithm is investigated. It is found that the performance of DLMS algorithm is related to the number of closed triangular row motifs in the network. Then, aiming at the problem of heavy communication burden between nodes in DLMS algorithm, a new algorithm of breaking mode-diffusion minimum mean-square algorithm is proposed. This algorithm greatly reduces the communication load between nodes, and the estimation performance loss of the algorithm is relatively small. Better balance of communication load and estimation performance is achieved, which plays an important role in saving network energy and bandwidth. This paper also applies diffusion strategy to phase estimation for the first time, and proposes a distributed phase estimation algorithm based on sensor networks combined with alternating iterative minimization method. The proposed algorithm can better resist noise interference. Then, from the overall structure of the network, this paper explores the influence of different network models in WSN, including regular network, small-world network, random network and scale-free network, on the performance of the proposed distributed phase estimation algorithm. It is found that the performance of the proposed algorithm is different with different network models. This study can be used to guide the topology design of sensor networks in distributed parameter estimation problems.
【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9;TN929.5

【参考文献】