节点文献

面向数据的无线传感器网络节能机制研究

Research on Data-Oriented Energy-Saving Mechanism in Wireless Sensor Networks

【作者】 黄如

【导师】 朱杰;

【作者基本信息】 上海交通大学 , 电子科学与技术, 2008, 博士

【摘要】 无线传感器网络作为一种新兴的、将改变人类与物理世界交互方式的新技术,具有广阔的应用前景和巨大的研究价值。无线传感器网络研究的重要目的是在满足网络面向应用要求的前提下最大化网络生命周期,由于能量资源约束是影响网络生命周期的最根本因素,因此,针对节能机制的研究在无线传感器网络研究领域中处于核心地位。本文从传感器网络的“数据中心和面向应用”的本质特征出发,以网络数据收集应用为背景,通过分析传感器网络中数据的内涵属性和分布特征,并结合网络能耗和结构模型,从降低和均衡网络能耗的角度研究能量有效的网络运作机制。首先,本文面向传感器网络中时间序列数据的时域关联性特征,针对网内时间冗余数据和流量不均衡分布模式所导致的传输能量浪费和漏斗效应问题,提出了基于预测模式的时间冗余数据滤波机制和能量感知数据路由机制。其中,时间冗余数据滤波机制的设计构架由捕获数据时域变化规律的预测模块、修正预测模型的数据自学习模块和控制数据滤波操作的驱动模块组成。设计中将预测精度阈值分配规则和预测误差驱动规则引入数据滤波体系的构造,通过针对节点能量状况个性化的预测精度阈值分配和根据预测误差精确判断所获得的数据变化规律内涵信息,进一步加强了针对时间冗余数据的识别和滤波效果。能量感知数据路由机制的设计结合了蚁群优化机理自适应网络状况动态性的优势和预测模型揭示数据流量变化规律的优势,通过将节点负载因子引入蚁群优化算法中启发式因子的构造和局部信息素更新规则的设计,赋予蚂蚁代理在路由解空间探索中预知网络局域能量状况的能力,提高了数据路由构建的自适应性和能量均衡性。实验表明,本文提出的面向时间序列数据预测模式的节能机制,通过挖掘数据内涵的时域冗余度和关联性,并引入蚁群优化机理与预测模式相结合的实现方式,有效地降低和均衡了数据收集能耗。其次,文章面向传感器网络中应用服务质量要求所体现的数据内涵价值特征,针对网内价值冗余数据传输造成的能量浪费和监控节点生命周期缩短的问题,研究了体现服务区分性的节能数据收集机制。文章分析了网络数据价值的分类判断方法并将其形式化为数据价值因子结构,进而在数据价值因子基础上,设计了映射为集合覆盖问题的价值贡献驱动节点调度机制和价值冗余数据滤波体系结构。其中,节点调度机制的设计思想是将体现价值贡献度的数据价值因子引入混合蚁群优化算法(IMAH)中启发式因子和全局信息素更新规则的设计,从而引导人工蚂蚁在解空间探索中的价值取向,在满足覆盖要求和能量有效的基础上通过迭代方式求取全局最优解。价值区分性数据滤波体系构建的节能思想是将数据价值因子引入支持QoS的MAC层退避机制的设计,通过控制网内不同价值含量数据包的发送优先级来实现减少网内价值冗余数据传输量的滤波效果。这种面向数据价值服务区分度的节能机制根据网络应用的服务质量要求优化节能效果,区别于传统的面向数据自身统计特征的节能处理方式。仿真实验表明,本文提出的面向数据价值的节能收集机制可根据不同级别的服务质量要求自适应地控制数据收集能耗,从而提供了从网络应用QoS的角度进一步优化机制能量有效性的新思路。然后,文章面向数据内容关联度特征,针对传感器网络数据收集应用中网内数据内容关联度低引发的聚类结构优化问题,和网内数据内容冗余度高造成的数据传输能耗浪费问题,提出了基于关联规则挖掘的聚类构建和结构优化算法,及内容冗余数据滤波机制。根据数据内容关联程度构建聚类结构,并通过簇重组和簇自愈算法从内容关联性角度进一步优化已有聚类结构,进而在聚类结构基础上,设计了针对数据内容冗余度特征的滤波算法,依据以“内容特征码”为核心的协商机制抑制内容冗余数据的传输,减少数据收集源头的数据产生量。并在已知数据内容关联度的基础上引入分布式信源编码方式来实现簇际传输数据的无损融合。本文提出的面向数据内容相关性和冗余度特征的节能机制设计,充分考虑了传感器网络数据收集实际应用中采样数据间内容相似度高的特性。实验结果表明,引入内容关联聚类和内容冗余滤波操作后,可以进一步降低数据收集机制的能耗。最后,文章面向传感器网络数据的分布统计特征,针对异构残缺数据的模型估计和分布模式规律挖掘的困难性,采用依据半监督学习估计的高斯混合模型描述异构数据的统计分布特征,并在数据分布模型的基础上设计了模型匹配度驱动的自适应数据滤波机制,该机制采用基于假设检验方法的模型匹配度判断来挖掘数据序列分布模式间的相似性,通过滤除冗余的分布式流数据序列,达到减少数据收集源头冗余数据产生量的目的,进而在簇际数据传输过程中,设计了基于主元分析的聚类数据压缩算法,通过冗余属性滤波和主元方向的数据重构,在传输数据降维的基础上实现满足累计方差贡献率的数据有损压缩,降低了数据传输过程中的能耗。实验表明,本文提出的基于异构数据统计特征的节能数据收集机制,有效解决了针对多属性混合和信息残缺性异构数据的建模和冗余度提取的难题,从面向统计特征的角度为传感器网络中异构数据的节能收集方法研究提供了新的设计思路。本文对所提出的方案和算法进行了充分的理论分析和实验验证,结果表明本文提出的节能机制能够从面向数据特征的角度进一步提高传感器网络运行机制的能量有效性,从而为传感器网络节能机制领域的研究提供了有益的探索。

【Abstract】 Wireless Sensor Networks as a novel technology, which could change the interactive mode between human being and physical world, has a wide application prospect and great research significance. The key purpose to study WSN is to maximize network lifetime by the premise of meeting QoS of networks. As energy resource constraint is the fundamental issue, thus the research on energy-saving mechanism is at the core position in WSN field. The dissertation focuses on essential characteristic of data-oriented and application-oriented in sensor networks, by analyzing the internal property, distribution of data, and combining energy consumption model, topologic structure of networks. For the aim of prolonging the lifetime of sensor networks, the dissertation studies energy efficient operation mechanism of sensor networks, through the research on reducing and balancing energy consumption.Firstly, the dissertation proposes prediction-mode-based filtering mechanism and energy-aware routing mechanism to solve the problems of waste of transmission energy cost and funnel effect respectively caused by time-redundant data and imbalanced flow distribution mode, according to the characteristic of temporal correlation on time series data in sensor networks. The design framework of filtering mechanism for time-redundant data is composed of prediction module for capturing the change law of time domain, self-learning module for updating model, and driving module for controlling data filtering operation. To build time-redundancy data filtering system, allocation rule on threshold of prediction accuracy and prediction-error-driven rule are introduced, personalized prediction threshold is allocated according to node energy status, internal information included in the data variation patterns is precisely judged and obtained based on prediction error bound, so as to further improve the recognition and filtering effect for time redundancy data. The design of energy-aware routing mechanism combines the advantages of ACO principle, which is self-adaptive to dynamic network situation, and the advantages of prediction module, which reveals the law of data flow change. By introducing node-load-factor into both construction of heuristic factor and design of local pheromone updating rule in ACO, artificial ant agents are endowed with perception ability of local energy status in WSN, and the self-adaptability and energy-cost-balance of routing construction are improved. The experiment result shows that, the above energy-saving mechanism effectively reduces and balances the energy cost of data gathering mechanism by mining the temporal redundancy and associability, and introducing ACO.Secondly, the dissertation studies energy-saving data gathering mechanism based on dipartite degree of service quality to solve the problems of energy waste and short life time of source node, according to QoS-oriented data value characteristic. The classified judgement methodology of data value is proposed, and formalized to the structure of data-value-factor. On the basis of data-value-factor, contribution-driven node scheduling mechanism which is mapped into SCP, and value redundancy data filtering system are designed. Contribution-driven node scheduling mechanism introduces data-value-factor into IMAH for the design of heuristic factor and global pheromone updating rule, which guides the artificial ant in solution space to obtain optimal solution based on value orientation, and further obtain the global optimization solution by iteration mode, on the premise of meeting the covering requirement. The design idea of value-diversity-based data filtering system is to transfer high-value packets with high priority and inhibit transmission of low-value packets by introducing data-value-factor into backoff mechanism in QoS-MAC layer, and finally reduces transmitted data amount, achieves filtering effect. The above energy-saving mechanism driven by QoS-oriented requirement is different from traditional modes with data statistical characteristic. The experiment result shows that, the proposed mechanism can adaptively adjust energy consumption according to different QoS levels, therefore, it is helpful to improve energy-saving effect.Thirdly, the dissertation proposes the methodologies for cluster construction and structure optimization based on mining data association rule, as well as content redundancy based filtering mechanism, to solve the problems of structure cluster’s optimization caused by content-low-correlation, and energy waste caused by content-highly-correlation. By using association rule to analyze the data content relevance, content-highly-correlation cluster is constructed, the existing structure of cluster through rebuilding and self-healing algorithm is further optimized. Furthermore, content-redundancy-based filtering algorithm is designed to filter the content-redundant transmitted data in cluster by building the negotiation mechanism which takes content characteristics code (CCC) as core. Then, according to known content-correlation, distributed source coding is explored for energy efficient lossless-data-fusion. The above energy-saving mechanism well considers the content-similarity-based universal phenomenon in WSN. The experiment result shows that, the energy cost is significantly decreased by introducing the content-highly-correlation cluster and the filtering operation on redundant data. Finally, according to the statistical-characteristic of data in WSN,GMM is adopted to describe the statistical distribution characteristics of heterogeneous and incomplete data using semi-supervised learning method, to solve the difficulty of building model for heterogeneous incomplete data and mining distribution-mode-law. Based on accurate data distribution model, adaptive filtering mechanism driven by model matching degree is proposed, it adopts hypothesis testing method to judge the similarity between different distribution patterns of data sequence, and reduces redundant data amount generated in the source of data gathering, in order to achieve the energy-saving aim by mining and filtering the redundant distributed flow data. Data compression algorithm based on cluster mode and principal component analysis (PCA) is proposed in external-cluster data communication process, by filtering the redundant attributes and data reconstruction based on PCA. On the premise of meeting the cumulative variance contribution rate, data dimensionality is reduced, and energy-saving effect is achieved. The above energy-saving mechanism effectively solved the modeling and redundancy extraction problem on multi-attribute and incomplete nature of heterogeneous data. From the statistical-trait-oriented view, the novel data gathering mechanism on incomplete heterogeneous data is proposed.In the dissertation, the efficiency of the proposed mechanisms and algorithms are proved by both theoretical analysis and simulation verification. Besides, the dissertation provides the helpful exploration to the data-characteristic-oriented energy-saving mechanism in WSN.

  • 【分类号】TP212.9;TN929.5
  • 【被引频次】4
  • 【下载频次】578
  • 攻读期成果
节点文献中: 

本文链接的文献网络图示:

本文的引文网络