【作者】 霍永凯；

【导师】 黄刘生；

【作者基本信息】 中国科学技术大学 ， 计算机软件与理论， 2009， 硕士

【摘要】 伴随着信息技术的发展,无线传感网络日益成为研究的热点,而信息收集和目标跟踪均是无线传感网络应用的关键技术。针对大规模移动无线传感网络数据收集中消息复杂度过高的问题,以移动目标跟踪为主要应用背景,我们开发了基于人的行为特征移动成簇算法(BMC)和基于概率的多簇快速信息统计算法(PCFD)。上述算法已成功的应用于“基于CNGI和WSN的矿山井下定位与应急联动系统”。在实际环境中,信号干扰将使节点在簇成员和独立节点间频繁的抖动,从而大大增加了成簇算法的消息复杂度。为解决该问题,我们设计了延迟保证的信号平滑机制(DBSS)和双阈值成簇机制(DTC)。我们还对信号平滑算法的参数选择问题进行了详细的理论分析。优化参数后的DBSS信号平滑算法可以去除信号中90%的波动,并削弱剩余毛刺50%的强度。DTC利用双阈值线来限制节点的加入和离开簇,阈值可根据系统需求调节。其可以有效的稳定成簇算法,进一步减小消息复杂度。在传统的跟踪系统中,移动节点周期性的发送定位信息包。随着移动节点数目的增加,这些方法将导致较高的丢包率,并缩短网络的生命周期。在实际应用中,我们发现很多移动节点彼此极为接近,以至于定位算法的精度无法区分,因此通过信息融合可以减少系统的消息负载。针对基于射频(RF)的人员跟踪系统,通过采用接收信号强度(RSSI),本文提出基于人的移动特征的成簇算法,根据移动节点间的距离进行成簇,并可以有效地维护簇。在BMC中,簇头代替了每个节点周期性地向网络发送定位信息包,因此极大的减小了消息复杂度。在簇维护中,我们设计了基于概率的状态通告机制(PSI),其可以高效地对簇进行维护。模拟显示基于BMC(采用了DBSS和DTC)的定位信息收集较传统的方法(TPTS)消息复杂度减小了64%。为快速高效的收集大规模移动网络中的信息,我们开发了基于概率的多簇快速信息统计算法。PCFD以概率的方法选举簇头,然后各簇头分别统计本簇内的消息,然后发送至网络。不但有效的控制了簇头数量,并可以根据信息相似度来组成簇,最大化消息融合度。实验及理论分析证明,该算法的消息复杂度可达O(M+n),其中M为网络节点数,n为簇头数量。本文的主要贡献和创新点如下:1.提出了基于权值和双曲线的信号平滑处理算法DBSS,并对其各参数选择进行了理论分析与优化。2.借鉴人的行为特征,设计了适用于人员跟踪的定位信息收集算法BMC,并提出了基于概率的状态通告机制PSI对簇进行高效的维护。3.开发了基于概率成簇的快速高效的节点信息收集算法PCFD。4.基于本文算法和理论在Micaz上开发实现了基于CNGI和WSN的矿山井下定位与应急联动系统。该项目为国家发改委下一代互联网重大专项。更多还原

【Abstract】 With the development of information technology, Wireless Sensor Network is increasingly focused. Both of data gathering and tracking are the key application of Wireless Sensor Network. Aiming at the problem of high message complexity in data gathering of the large scale mobile Wireless Sensor Network, taking tracking as our main applications, we develop human-Behavior based Mobile Clustering Mechanism (BMC) and Probability-based multi-Cluster for Fast Data gathering (PCFD) for large-scale mobile network. The above algorithms are successfully applied in the pro-ject“CNGI and WSN Based Mine Underground Localization and Integrated Emer-gency Response System”.Signal interference will wobble the clusters in practical environment, so it de-grades the performance of the clustering algorithm dramatically. To resolve the issue, Delay-ensured Best-effort Signal Smoothness (DBSS) and Dual Threshold Clustering (DTC) are developed. We also theoretically analyze the selection rules of the smoothness parameters. With the optimized parameters, DBSS smoothes out more than 90% of the fluctuations and the stubborn ones left are also weakened by about 50%. DTC adopts two thresholds lines to guard joining and leaving of a cluster. The thresholds can be adjusted according to the requirement of the system. It can reduce the message complexity greatly, thereby stabilizing the clustering algorithm.In traditional tracking systems, the mobiles report their locations periodically. With the number of the mobiles increasing, these methods will result in high loss rate of packets and rapid depletion of the network energy. In practice, we observe that some mobiles are so close to each other that we are informed the same location from the localization algorithm. So it is necessary and possible to reduce message complex-ity through merging the location messages. By exploiting the Received Signal Strength Indicator (RSSI), this paper proposes a human-Behavior based Mobile Clus-tering Mechanism for Radio Frequency (RF)-based Person Tracking Systems. It con-structs clusters according to the distance between mobiles and maintains clusters effi-ciently. In BMC, only the cluster-heads report locations periodically instead of each node, thus the message complexity is reduced greatly. In the cluster maintenance stage, we devise the named Probability-based State Informing mechanism (PSI), which can maintain the clusters effectively. The simulation shows that BMC based location re- port outperforms traditional ones by 64% of message reduction on average in impar-tial testing scenes.To gathering the data quickly and effectively in large-scale mobile Wireless Sensor Network, we develop Probability-based multi-Cluster for Fast Data gathering. PCFD selects cluster head with a variable probability. Then every cluster collects its own message and sends to the network. This method can not only control the number of clusters though adjusting the probability is that, but also construct the clusters ac-cording to the interests. Both simulation and theoretical analysis prove that the mes-sage complexity is O(M+n), where M is the scale of the network and n is the number of clusters.The contributions of this dissertation are listed as follows:1. We propose a signal smoothness algorithm DBSS, which is based on weighted means and dual curves. Also we analyze and optimize the selection rules of DBSS.2. Exploiting the traits of human behavior, we design a location data gathering algo-rithm for person tracking and the probability based state informing mechanism (PSI) is devised to maintain the cluster effectively.3. The probability based clustering algorithm PCFD is proposed to gathering the nodes’information quickly and effectively.4. Based on the algorithm and theory proposed in this dissertation, we develop and implement the CNGI and WSN Based Mine Underground Localization and Inte-grated Emergency Response System on the Micaz platform. The project is China the Next Generation Internet Project Important Special Item from the State De-velopment and Reform Commission of China.更多还原