【作者】 孔凡天；

【导师】 周祖德； 陈幼平；

【作者基本信息】 华中科技大学 ， 机械电子工程， 2006， 博士

【摘要】 无线传感器网络具有布线成本低、监测精度高、容错性好、可远程监控、便于诊断与维护等众多优点,在网络制造、智能制造等领域有着广阔的应用前景,其根本任务是准确获取物理世界中有价值的信息。无线传感器网络借助节点的时间与位置信息,实现传感器节点之间控制和传感数据的高速率、低延迟交换,以保证整个检测与控制系统的准确性与实时性。然而,无线传感器网络面临着计算、存储与网络资源等方面的限制,针对无线传感器网络的体系结构、节点时钟同步、节点定位以及数据融合等方面的问题展开研究具有十分重要的意义。首先,本文详细研究和讨论了无线传感器网络的体系结构,提出了一种新颖的无线传感器网络分层式结构模型。传统的网络体系结构是为各种应用程序提供网络传输上的支持,是以实现“传输为目的”的,节点只具有传输功能,不对数据进行处理。而无线传感器网络“以数据为中心”,其目的是获取被感知对象的长期、准确的特征信息。该体系结构的设计遵循节约能耗的原则,尽可能地对传统网络中的非主要功能进行了约简,将OSI参考模型中的网络层和传输层合并为网络层,同时在协议中应用多传感器数据融合技术达到了节省存储资源、降低网络带宽占用的目的。分析了无线传感器网络的节点时钟同步机制,提出了一种基于时间戳的节点时钟同步算法。基于时间戳的节点时钟同步算法的一个重要特征是参照时钟不同步的传感器节点的时间信标对时间戳进行转换从而实现时钟同步,而不直接进行节点之间的同步。由于时间戳信息可以通过现有的消息交换携载,该算法具有能源高效性。实验结果表明基于时间戳的节点时钟同步算法可以获得毫秒级的同步精度,已经可以满足无线传感器网络的大多数实际应用。深入讨论了无线传感器网络节点定位算法的基本原理,提出了一种适用于节点具有移动性的无线传感器网络的基于蒙特卡罗方法的节点定位算法(Monte Carlo Localization, MCL)。MCL算法在网络节点(包括信标节点和普通节点)的运动不可控的情况下,不需要额外的硬件设施,就可以获得比现有几种定位方法更高的精度,解决了具有移动性的无线传感器网络中的节点定位问题。研究了无线传感器网络中传感数据的融合处理方法,提出了一种基于无线传感器网络的分布式K-平均聚类算法(Distributed K-mean Clustering method for Wireless Sensor Networks, DKCWSN)和一种基于自适应加权的网内数据融合方法。DKCWSN算法可以实现无线传感器网络节点传感数据的快速、合理分组,而基于自适应加权的数据融合方法可以更好地将分组后的节点传感数据按其权重值的大小,作出正确的判断,融合处理得到更合理的结果。该方法简单、实用,大大降低了整个网络内的数据冗余度,节省了大量的存储资源和网络带宽。算法己在原型系统中实现,实验结果表明它具有较高的执行效率和很好的可扩展性。最后,基于上述关键技术与方法,设计了一种基于无线传感器网络节点位置信息的移动机器人定位导航系统。同时,为了适应远程测控、远程医疗、公共交通、公共信息查询等领域的应用,对基于ZigBee协议的无线传感器网络进行了分析与研究,开发了一种基于该无线传感器网络的粮仓测控系统。更多还原

【Abstract】 The primary advantages of wireless sensor networks include reduction of system wiring, high monitoring accuracy, convenience for remote supervisory controlling, easiness of system diagnosis and maintenance, good fault-tolerant ability, and etc. There is wide application prospect of wireless sensor networks in such fields as networked manufacturing, intelligent manufacturing and so on. The basic task of wireless sensor networks is to obtain valuable information of the physical world accurately. Wireless sensor networks can realize high speed and low time-delay message exchange among sensor nodes depending on the time and position information. So it ensures the high precision and real-time characteristic of the whole system. Wireless sensor networks face many-sided restrictions such as calculation, storage and network bandwidth. It is very important for wireless sensor networks to investigate the system architecture, clock synchronization, node localization, data fusion, and etc.First of all, the system architecture and related theories of wireless sensor networks were studied and discussed in detail. Then a kind of novel system hierarchical model of wireless sensor networks was proposed. The traditional network system architecture offers the network transmission support for various kinds of application program with the purpose of data transmission.The network node only has a transmission function, and it doesn’t need to deal with any data exchange. On the contrary, wireless sensor networks is focus on sensing data, of which the aim is to obtain more accurate characteristic information of the perceived targets for a long time. Design of this system architecture followed the principle of energy conservation. It has simplified the non-main function in the traditional network as much as possible, and it has merged the transmission layers of OSI model into the network layer. At the same time, data fusion technology was also used in the networks protocol so as to achieve the goal of saving storage resources and reducing network bandwidth occupation.Clock synchronization mechanism of sensor nodes in wireless sensor networks was analysed, a kind of time synchronization algorithm based on time-stamp was presented. The main characteristic of the algorithm based on time-stamp was to make time-stamp transformation according to the asynchronous clock sensor nodes’time-scale. And it didn’t carry out synchronization among the nodes directly. Since the time-stamp information can be taken by existing exchanged messages, this method was energy-effective. The experiment result indicated the proposed algorithm was able to obtain a synchronization precision of one millisecond in order of magnitude, which could satisfy most of the practical applications of wireless sensor networks.The basic principles of node localization scheme for wireless sensor networks were discussed thoroughly. Then a new kind of localization algorithm based on Monte Carlo method was proposed, which was suitable for wireless sensor networks with node mobility (Monte Carlo Localization, MCL). MCL algorithm can obtain much higher precision than several other traditional localization methods under condition of the sensor node’s movement uncontrollable. The method didn’t need any extra hardware facilities, and it could solve node’s localization problem for wireless sensor networks with mobility.The data fusion technology in wireless sensor networks was studied deeply, a kind of Distributed K-mean Clustering algorithm for Wireless Sensor Networks (DKCWSN) and data fusion algorithm based on self-adapting weight were presented. DKCWSN algorithm could realize sensor nodes’fast clustering throughout the whole network. After the sensor nodes were divided into many groups reasonably, the simple data fusion algorithm based on self-adapting weight could make correct decision, and it was able to deal with the received sensing data to obtain more reasonable result according to the corresponding weight size. This method is simple, practical, and the redundant degree of data in the network could be decreased to a great extent. Samely, it could save storage resources and bandwidth occupation of the network. The algorithms had been implemented among the prototype system, and experiment result showed that the algorithms had good execution efficiency and very kind expansibility.In the end, on the basis of the methods mentioned above, a mobile robot navigation prototype system using wireless sensor networks node position information was realized. The wireless sensor networks with ZigBee protocol had been studied in detail, and an embedded granary measuring and controlling system using this kind of ZigBee wireless sensor networks was developed. The wireless sensor networks based on ZigBee protocol could be widely used in such fields as long-range measuring and controlling system, telemedicine, public traffic, public information inquiry, and etc.更多还原