【作者】 郑瑾；

【导师】 贾维嘉；

【作者基本信息】 中南大学 ， 计算机应用技术， 2010， 博士

【摘要】 无线传感器网络是以数据为中心的网络,感知数据的管理与处理是实现高效率传感器网络的关键任务之一。与传统无线网络相比,传感器节点的能量非常有限,而且常常由于特殊的工作环境而无法得到补充,网络生命期直接受到传感器节点能量的影响；同时,无线传感器网络的数据管理与传统的分布式数据库存在很大的差别,因此,必须针对传感器网络的特点,设计能量高效的、可靠的、简单易实现的数据管理算法。由于传感器节点本身的能量、存储与计算能力非常有限,现有复杂的无线传感器数据管理优化技术难以在大规模无线传感器网络中得到实现,从而导致数据管理技术的研究结果距离真正的实际应用还有较大的差距。论文以有效管理无线传感器网络中的数据为目标,针对无线传感器网络的内在特点和已有研究的不足,研究了无线传感器网络中分布式数据管理的一些关键技术。论文的贡献主要集中体现在以下四个方面：数据收集针对需要收集网络产生的全部数据的数据收集问题,提出基于移动Sink自适应移动的能量高效及能量均衡的数据收集协议(EEBDG)。协议采用基于贪心策略的移动决策和边反向技术,Sink总是向数据流量大或剩余能量小的区域移动。主要解决传统的Sink移动需向网络更新其位置信息而存在位置更新代价高的问题。EEBDG从能量高效和能量均衡两方面出发延长网络生命期,适用于在传感器节点存储容量有限、网络数据流量分布动态变化、数据收集延迟敏感的大规模无线传感器网络的数据收集。(?) Top-k查询针对无线传感器网络的基础查询问题--Top-k查询,即用户只需要网络返回在某种比较规则条件下的前k个数据的查询问题,提出基于数据分布表的查询协议(DDT-Q)。采用跨层优化的策略,利用数据分布表,将查询请求只分发到对查询结果有影响的节点,避免查询请求在网络内部泛洪从而减少查询分发代价；减少对查询结果无影响的数据在网络内的传输从而减少数据传输代价；根据返回数据量的大小分配带宽从而减少查询延迟和能量消耗。DDT-Q协议简单,能够在大规模无线传感器网络中部署实现。支持目标轨迹查询的分段链式存储和查询协议针对没有Sink、面向目标跟踪应用的无线传感网络中目标及目标轨迹查询问题,提出分段链式存储和查询协议(SLDSQ).采用本地存储和以数据为中心存储及协作存储的存储策略,将跟踪数据存储在其产生附近的节点中,并以链的形式保存存储节点之间的关系,采用分段式的索引模式,在能量高效的前提下,有效地支持用户在低延迟下得到目标或目标轨迹的查询结果。多目标环境下的目标及目标轨迹存储和查询协议对于目标较多的跟踪应用中的目标及目标轨迹查询问题,提出随机查询和索引查询相结合的查询策略(CRIQ)。基于目标进入监测区域存在时间和空间上的相关性,利用“搭乘”思想,即利用跟踪过程本身的通信机会,将目标信息在无额外通信代价的情况下在网络内散发实现多点存储,从而降低随机查询的代价。同时设置分布式的索引节点,记录目标及目标轨迹索引信息。随机查询和索引查询相结合的查询策略具有能量消耗低、查询延迟低、查询结果可靠性高和避免索引节点的查询热点等特点。更多还原

【Abstract】 Wireless Sensor Networks (WSNs) are data-centric networks. It is one of their key missions to manage and process sensed data for WSNs with higher efficiency. Compared with traditional wireless networks, nodes in WSN are very energy constrained and often not rechargeable due to its special working environment, which directly affect the lifetime of the whole network. Meanwhile, data management in WSN is very different from that in traditional distributed databases. Therefore, we must design an energy efficient, reliable and simple data management algorithm fitting the features of WSN nodes. However, current complicated WSN data management optimization technologies can not be implemented in large scale WSNs, because sensor nodes are limited in battery energy, storage and computation capacity. As a result, there is a large gap between researches of data management in WSN and real implementations.Aiming at the inherent characteristics of wireless sensor networks and the limitation of current work, this dissertation takes the efficient management of data in wireless sensor networks as the goal and studies some key technologies of distributed data management of wireless sensor network comprehensively. The main contributions of this dissertation can be summarized in the following four aspects.(?) Data CollectionRegarding the problem of collecting all data generated by the network, we propose an energy efficient and energy balancing of data gathering protocol based on adaptive Sink node movement(EEBDG). The protocol leverages a movement strategy based on greedy algorithm and the side reversing technique, so that the Sink node always moves towards areas with large data flow or little remaining energy. We made a breakthrough by solving the problem of too large position update overhead in traditional WSNs that require the Sink node to broadcast its latest position. EEBDG can actually prolong the lifetime of the WSN networks by energy saving and energy balancing. It works best with WSNs which have nodes limited in storage and are sensitive to data collection delay.(?) Top-k QueryRegarding the basic query problem in WSN, i.e., Top-k query, where users only need the first k data generated by the network under some rule, we propose Distributed Data Table Query (DDT-Q) strategy. Based on cross-layer optimization policy, DDT-Q makes use of data distribution table to disseminate queries only to the nodes which may influence the final results and avoids flooding the network with query requests. Thus the query dissemination overhead is reduced. Meanwhile, the data transmission overhead is also reduced, because we avoid transmitting data which will not affect the final results. The communication bandwidth is assigned according to the transmission data volume, which leads to less delay and energy consumption. DDT-Q is simple and easy to implement, and can be deployed in large-scale WSNs.(?) Segmented Linked Data Storage and Query ProtocolAs for the problem of target trajectory query in target tracking oriented WSNs without Sink nodes, we propose Segmented Linked Data Storage and Query (SLDSQ) protocol to support user segmented trajectory query mode. SLDSQ uses a storage strategy combining local storage, data-centric storage and colabrative storage. The tracking data are stored in the nodes generating the tracking data. The relationship among storage nodes is stored as a chain. Segmented indexing storage mode is used to support multi-user queries of targets and trajectories without much delay.(?) Querying Protocol in Multi-Target Tracking ApplicationsTo solve the problem of target and trajectory query in multi-target tracking applications, we propose the querying protocol CRIQ that integrate random query and index query. Based on the temporal and spatial correlation of targets’ entrance into the monitored area, piggybacking during exchanges of tracking information is used to diffuse the target information, implement multi-point memory in the WSN without extra overhead and, thus, reduce random query overhead. Meanwhile we appoint distributed indexing nodes, which register the information of targets and their trajectories. The query strategy integrating random query and index query can reduce energy consumption and query delay, as well as to ensure the reliability of query results and avoid query hot spot problems of indexing nodes.更多还原