【作者】 李湃；

【导师】 陈向东；

【作者基本信息】 西南交通大学 ， 通信与信息系统， 2007， 硕士

【摘要】 无线传感器网络(WSN)是由大量低成本且具有传感、数据处理和无线通信能力的传感器结点自组织构成的多跳网络。可以在大范围内用于收集、处理和发布复杂的环境数据。借助传感器网络执行一些具体任务时，任务完成质量是由网络性能决定的，如网络生存时间、网络负载平衡性、网络稳定性等等。这些性能都与网络结构密切相关。无线传感器网络大多采用分簇结构。网络建立时，通过分簇算法将网络划分为若干个簇，每个簇由一个簇头节点和若干簇成员节点组成。簇头负责管理簇内资源分配和簇之间通信。簇成员节点负责数据的采集和处理等。因为簇结构对网络性能有很大影响，所以研究无线传感器网络分簇算法对提高网络性能有重要意义，因此在实际应用中也具有重要意义。这篇论文对无线传感器网络分簇算法进行研究，主要工作如下：一、将无线自组网络中三种分簇算法引入无线传感器网络：其中一种自适应按需加权分簇算法比较复杂，针对这一点，使用分项阈值判别代替分项数值计算，提出一种新型的自适应按需加权算法。新算法的特点是，运算简单，性能良好，负载平衡性突出。二、现有分簇算法一般只选取一个簇头，有簇头消亡便要重建网络，导致网络稳定性不好。现提出一种新型“双簇头”分簇机制，即分簇时产生两个簇头，一个做簇头，一个做备用簇头，原簇头消亡时，备用簇头升为簇头，无需重建网络，从而降低网络更新频率，提高了稳定性。这种新机制可以引入现有的一些分簇算法当中。引入新机制后，网络的稳定性比原算法有很大提高，节点充当簇头的公平性也有一定提高。更多还原

【Abstract】 Wireless sensor network (WSN) is multiple-hop network self-organized with lots of low-cost sensors with ablities of sensing, data processing and wireless communication. Wireless sensor network configured automatically is highly valuable in both civil aspect and military aspect. Carrying out some tasks, the accomplishment of task depends on the network performance, such as life time of network, the balance of load and stability etc. All of the capabilities relates to network hiberarchy. Hiberarchy is mostly adopted in WSN. As WSN is established, the network is divided into some clusters. Every cluster is made up of a cluster head(CH) node and some cluster members. The cluster head manages to distribute resource in cluster and communication of clusters. Cluster members are responsible for data collecting and processing. Because clustering has great influence on network capability, the research on clustering algorithms in WSN is very important for the advance of network performance, in practical application as well. The thesis expatiates the research aiming at some clustering algorithms in Wireless Sensor Network, and the central achievements is shown as follows:1. In the paper three clustering algorithms in Ad Hoc network are introduced into wireless sensor network. One of these algorithms, which called Automatic On-demand Weighted clustering algorithm, is complex. In order to reduce the complexity of that one, a new algorithm is presented which adopted threshold estimation , instead of numerical value account. The new algorithm indicates the simpler operation, the better performance, and better load balance.2.Current clustering algorithms of network only select one cluster head in a cluster. When the cluster head dies out, the network will be reestablished, and stability of the network will be reduced. So a new " Double Cluster Heads " clustering mechanism is put forward. The mechanism brings two cluster heads: a formal cluster head;the other is spare cluster head.When the first dies out, the second will become formal cluster head. It dispenses with network re-establishment, reduces network updating frequency, and improves network stability. The mechanism may be introduce into existing algorithms. Compared with old algorithms, both stability and Head Fairness of wireless sensor network have been greatly improved after the new mechanism is introduced into existing clustering algorithms.更多还原