节点文献
面向动态环境监测的无线传感器网络数据处理方法研究
Data Processing Method of Dynamic Environmental Monitoring for Wireless Sensor Network
【作者】 吴秋云;
【导师】 景宁;
【作者基本信息】 国防科学技术大学 , 信息与通信工程, 2013, 博士
【摘要】 微电机系统、传感器、无线通信和低功耗嵌入式技术的飞速发展,推动了现代无线传感器网络产生和发展,拓展了对信息的感知和获取能力,并以其低功耗、低成本、分布式和自组织的特点带来了信息感知的一场变革。无线传感器网络是涉及多个学科交叉知识高度集成的和得到学术界非常关注的研究前沿领域。目前,无线传感器网络中的一些关键技术内容仍然需要深入研究,比如:数据管理和数据安全、能源管理、服务质量、负载均衡等问题。必须将这些关键技术问题攻克了,才能在实际应用中真正发挥出无线传感器网络潜在的巨大作用。由于无线传感器网络往往部署在网络结构动态变化、数据来源不确定等具有各种复杂因素的动态环境中,如何有效进行数据处理具有很大挑战性。本文从面向动态环境监测的角度对无线传感器网络的数据处理方法进行研究,主要研究工作包括:对面向动态环境监测相关关键技术研究及其应用现状进行了总结。探讨了面向动态环境监测的无线传感器网络的概念、特点和应用前景,分析了面向动态环境监测的无线传感器网络数据处理所面临的挑战。研究了面向动态环境监测的缺失数据估计方法。在面向动态环境监测的情形下,感知数据的缺失问题给无线传感器网络的各种应用带来了巨大困难,不仅降低感知数据集合的可用性,而且使感知数据集的利用率急剧下降,还间接地降低了无线传感器网络的工作效率。本文基于物理位置上相邻的传感器节点采集到的监测数据往往比较相似或存在某种函数关系的特点,提出了基于时空自然最近邻的缺失数据估计算法STNNI,该算法估计准确性、稳定性好。研究了面向动态环境监测的无需测距定位方法。许多无线传感器网络应用中的感知数据必须和位置信息关联才有意义,获取无线传感器网络节点位置信息的定位技术是必须解决的关键技术。在分析无线传感器网络自身定位系统和算法的分类的基础上,提出了一种无需测距定位算法NDV-Hop。在该方法中,首先建立以信标节点为原点的跳数梯度场,得到待定位节点到信标节点的跳数距离;然后利用待定位节点到最近邻信标节点的近似距离和最近邻信标节点到其它信标节点实际距离,逼近待定位节点到各信标节点实际距离,从而优化待定位节点至各信标节点的平均距离值,减小了累加误差,提高了定位精度。该算法简单易行,无需额外添加硬件,可以满足一定应用需求。研究了面向动态环境监测的事件检测方法。事件检测是无线传感网络的一种主要任务,无线传感器网络一般部署在恶劣的环境中,针对无线传感器节点的软故障会产生错误数据,这些错误数据会降低事件检测算法的精度和性能,甚至产生虚报事件的情况,提出了一种基于自然近邻统计的事件边界检测分布式算法NNB-DEBD。节点只需要和自然邻节点交换一次所采集的感知信息,就可迅速检测该节点是正常节点还是故障节点,当正常传感器节点所感知的监测值达到事件触发的阀值条件时,可通过邻域统计的方法判断传感器节点是否处于事件发生边界上,事件边界宽度可依据实际应用需求进行调整。该检测算法时延小、复杂度低,算法所需通信的信息量小,有很好的可扩展性和稳定性,能适应于检测大规模无线传感器网络的事件边界。针对空间事件检测的特殊性,建立了空间事件模型,在该模型基础上扩展定义了空间事件复合算子及其语义,并证明了该定义的复合算子是封闭的;基于事件公共表达式的简化Petri网,构造了一个复合事件检测模型,经设置变迁优先级克服了冲突变迁的问题,针对该模型提出了一个检测算法,应用仿真实验验证了该检测模型的有效性。研究了面向动态环境监测的数据聚合方法。节省能量和延长网络的生命周期是动态环境下无线传感器网络面临的一个重要问题,对网内信息处理算法的适应性和鲁棒性提出了很高要求。网内聚集机制作为一种高效、节能的数据聚集方式,可以充分利用节点的自身处理能力对大量的冗余数据进行网内处理,在中间节点转发原始数据之前就对数据的聚集合并,消除冗余信息,合理权衡感知数据的精度和能耗,在满足实际应用需求的前提下尽量减小网络通信量,减轻网络拥塞,降低能耗,延长网络寿命。针对网内聚集机制提出一种基于近似最小生成树聚合算法GLB-MST,该算法复杂度较低且具有较好的节能效果。
【Abstract】 Wireless sensor networks is integrated with technologies such as sensor networks,embedded computing, distributed information processing, and wireless communicationtechnologies. It can not only collaborately monitor, sense and collect information of avariety of environmental monitoring object, but also process and transmit the data.Wireless sensor networks is a new interdisciplinary research field with broad applicationprospects which is causing a high degree of attention both in academia and industry.However, a number of open problems within the research scope of wireless sensornetworks are still at the exploratory stage, such as energy management, data managementand data security, QoS guarantee, opportunistic routing and other issues. Only after solvingthese technical issues, wireless sensor networks can really play a potentially huge role. Thewireless sensor networks is often deployed in dynamic environment with dynamic changesof network structure, data sources of uncertaint as well as other complex factors. How toeffectively carry out data processing becomes a challenging task. The dissertationinvestigates data processing methods of wireless sensor networks from the point of viewfor dynamic environmental monitoring. The main research work can be summarized as thefollowing five aspects.The key technology, its application and the state of art are summarized. The concept,characteristics and prospects of wireless sensor networks for dynamic environmentalmonitoring are investigated in detail. Besides, the main challenges of wireless sensornetwork data processing for dynamic environmental monitoring are analyzed.Estimation of missing data for dynamic environmental monitoring are presented.Under the condition of dynamic environmental monitoring, the lack of perceived databrings great difficulties to effective applications of wireless sensor network. It reduces boththe availability and utilization ratio of the sensing dataset. Moreover, such problemindirectly reduces the overall efficiency of wireless sensor networks. It turns out that themonitoring data collected by physically adjacent sensor nodes are usually have somesimilarities or functional relations. Taking such characteristics into consideration, a highlyaccurate and stable algorithm for estimating the missing data based on the spatial andtemporal natural neighbor (STNNI) is proposed.A novel localization algorithm that is range-free for dynamic environmentalmonitoring is presented. In most wireless sensor network applications, the perception databecomes valuable only when associated with location information. Therefore, it isnecessary to address how to efficiently position a wireless sensor network node. Based onthe analysis of wireless sensor network self-positioning systems and algorithms, arange-free positioning method is proposed. To position a node, a gradient field of thebeacon node as the origin of hops is created in the first step where the distance in hops from the node to the beacon node can be obtained. Then, the real distance from the node toall the beacon nodes can be approximated by the approximate distances between the nodeand the nearest neighbor beacon node, as well as the real distances between beacon nodes.By applying the positioning method, the estimation of the average distance between thenode to be positioned and the beacon nodes can be optimized. The accumulative error isminimized, and the positioning accuracy is improved. The algorithm is so concise andeasy-to-implement that no extra hardware is needed. So it is fairly practical in realapplications.Some Event detection methods for dynamic environmental monitoring are introducted.Event detection is one of the main tasks of wireless sensor networks. As wireless sensornetworks are generally deployed in severe environments, soft failure may cause the nodesto provide erroneous data, which reduces the monitoring accuracy, and results in thefalse-alarm problem. To cope with the problem, a distributed fault-tolerant event boundarydetection algorithm based on neighborhood statistics (NNB-DEBD) is proposed. Whenapplying NNB-DEBD, the soft failure of a node can be fast detected by exchanging thesensing data with its neighbor only once. When the monitoring value of a normal nodetriggers the event condition, neighborhood statistics is applied to determine if it is on theevent boundary whose border width can be adjusted according to the user requirements.The communication volume required by the algorithm at runtime is low. In addition, thealgorithm has low computational complexity, latency and a good scalability for large-scalenetworks. To depict the specifics of the spatial event detection, a spatial event model isdeveloped. Based on the model, the composite operators and their semantics of spatialevents are defined. The closing property of the composite operators under the definition isproved. Composite event detection model is established on the basis of colored Petri netswhich can be simplified by utilizing the event public expression. The problem ofconflicting transitions can be solved by applying their priority changes. A detectionalgorithm based on the model is proposed. And an experimental simulation of the detectionis adopted to verify the feasibility of the model and algorithm.Data aggregation method for dynamic environmental monitoring is presented. To savethe energy and prolong the network lifetime is very important to wireless sensor networksin a dynamic environment. Such requirement asks the information processing algorithmswithin the network be highly adaptable and robust. Aggregation mechanism within thenetwork is an effective and energy-efficient data aggregation strategy, which can take fulladvantage of the sensor node’s own processing capability. When transmitting the raw datato the base station, the data can be aggregated at the intermediate nodes. The energyconsumed by the network can be significantly reduced through a reasonable trade-offbetween accuracy and energy consumption. An approximate aggregate minimum spanningtree algorithm GLB-MST for in-network aggregation mechanism is proposed, which has a low computing complexity and good energy saving performance in practice.
【Key words】 Wireless sensor network; data processing; self-positioning; data aggregation; dynamic environmental monitoring;