【作者】 刘逵；

【导师】 刘三阳；

【作者基本信息】 西安电子科技大学 ， 应用数学， 2013， 博士

【摘要】 无线传感器网络是由部署在监测区域内大量的廉价微型传感器节点通过无线通信方式形成的一个多跳自组织网络，其可广泛应用于军事侦察、环境监测、城市交通管理和仓储管理等领域。无线传感器网络作为当今信息领域新的研究热点，涉及到多学科交叉的研究领域，因此有非常多的关键技术有待研究与开发，其中网络优化技术和动态组网技术是急需研究开发的核心问题之一。为了解决该问题，并进一步增强网络的自适应性和鲁棒性，本文对无线传感器网络优化技术和动态组网技术进行了相应的研究，所取得的主要研究成果如下：1.针对无线传感器网络的拓扑结构动态变化比较频繁、无线通信链路不稳定及节点能量有限等特性，本文提出了一种基于感染球策略的移动代理能量有效路由算法。该算法首先赋予移动代理以蚂蚁的特性，从而提高了算法的自适应能力；然后该算法利用最大感染球来缩小移动代理寻找及修复最优路径的范围，进而有效降低了网络的寻路能耗。同时本文还提出了一种最优路径局部快速修复策略，该策略能在保留原有最优路径大部分信息的基础上，对最优路径作出局部快速修复。与其它算法相比，本文所提算法能找到一条均衡各节点剩余能量和路由能量总消耗的最优路径，并能在寻路过程中避开剩余能量少的节点，使网络中各节点的能量呈整体性衰落，进而延长网络寿命。2.无线传感器网络中节点重要性评估的有效性，是优化网络拓扑结构和增强网络抗毁性的基础。现有的节点重要性评估方法大多是基于复杂网络中心化理论设计的，这类评估方法未考虑网络成簇等现象对节点重要性的影响，因此该类评估方法不适合于无线传感器网络。针对该问题，本文基于谱分析理论提出了一种簇收缩策略的节点重要性评估算法。该算法首先利用网络的非平凡特征向量来获取传感器网络的原始簇结构；然后借助模块度的增量来评估合并这些原始簇，从而形成一个与真实网络相匹配的簇结构；最后利用本文设计的簇收缩策略提炼骨干网络，并逐一评价骨干网络中的关节点。同时通过在重要关节点的一跳通信范围内投放超级节点，来完成对重要关节点的针对性保护，进而达到优化网络拓扑结构和增强网络抗毁性的目标。3.研究了单一类型基站在无线传感器网络应用中出现的弊端，提出了一种混合基站策略的网络数据收集算法。该算法首先在传感器网络中引入了两种不同类型的基站,进而有效解决了由固定基站带来的网络路由空洞问题和由移动基站带来的通信延迟问题；其次该算法采用了一种中途数据拦截策略，该策略可有效减少网络中采集数据的传输距离，进而提高网络数据的安全性；最后该算法还采用了节点拥挤自适应策略，该策略可根据节点剩余缓存空间的大小自动预警节点的拥挤繁忙程度，并及时利用网络中的备用路径来分流途经拥挤节点的传输业务，进而有效降低拥挤节点的繁忙程度，提升数据传输的正确率。4.针对高负载无线传感器网络堵塞率比较高的问题，本文提出了一种基于双信道策略的传感器网络联合优化路由算法。该算法首先利用双信道通信模式降低信道竞争过程中的数据碰撞和多播抑制几率；然后再利用最大感染球策略来压缩蚁群的寻路范围，进而降低网络的寻路能耗；最后该文借助分层图模型提出了一种双层网络联合优化的选路策略，该策略可将控制层中被堵塞的寻路业务有条件的下放到数据层中传输，从而降低网络的堵塞率和通信延迟。仿真结果表明，与其它算法相比，本文所提的算法能将高负载网络下的堵塞率下调13%，且能有效降低网络中数据包的平均传递时间和通信能耗。5.针对无线传感器网络中节点间距离存在测量误差的问题，本文在局部最小生成树算法的基础上借助鲁棒离散优化理论提出了鲁棒最小生成树算法。当不确定环境对最小生成树优化模型中的目标参数产生干扰时，则寻找无线传感器网络中的最小生成树问题就可借助布尔规划模型转化成一个离散优化问题。同时对于目标参数不确定的布尔规划模型，鲁棒离散优化理论证明了求解该模型的鲁棒对应可以转化成求解一个确定性的规划问题。因而当网络中节点间的测量距离均存在测量误差时，借助本文所提算法只需求解一个确定问题就可找到网络的鲁棒最小生成树。实验显示，无论监测区域内实验条件如何变化，本文所提算法找到的最小生成树上的节点总是保持有更高的度数，进而保证了环境恶化时最小生成树的抗毁性和鲁棒性。更多还原

【Abstract】 Wireless sensor networks is a multi-hop self-organization networks constructed bya large number of low-cost micro-sensors deployed at the interior of detection regions.The networks is formed by wireless communication and widely used in manyapplications, such as battlefield surveillance, environment monitoring, urban traffic andwarehouse managements, etc. Wireless sensor networks as a new hot research area inthe field of information involve in the realms of wireless communication and digitalelectronics. Therefore there are many key technologies to be studied. Networkoptimization technology is one of the core issues need to be considered. To solve thisproblem and improve robustness of networks, this paper studies networks optimizationand dynamic networking technology, and presents some effective methods. The workhas been done includes five facets are listed as follows:1. According to networks characteristics like frequent changes of topologystructure, unstable of the wireless communication line, limited node energy, etc., anenergy efficient mobile agent routing algorithm (EEMAA) base on the infection sphereis presented in this paper. Mobile agent is conferred the character of ant in thisalgorithm, and the infection sphere is used to reduce the number of nodes which join inresearching and restoring the energy efficient route from processing node to target nodes.These ways can reduce energy consumer of networks in searching for the optimal route.Meanwhile, a new restore rule for the failure optimal route is presented. By using thisrule, the optimal route can restore quickly in the local of fail nodes and most of theinformation of original optimal route can be reserved. Simulation results show that ourapproach can keep away from the nodes with less residual energy and make the energyof each nodes on the optimal route overall decline. Hence improve the lifetime ofnetworks.2. Effective evaluating node importance in wireless sensor networks is importantto the optimization of topology and the reliability of networks. The existing evaluatemethods are based on center theory of complex networks, which don’t consider theinfluence of cluster in networks. Hence, they don’t suitable for wireless sensor networks.In this paper, based on agglomeration contraction principle for wireless sensor networks,we propose a novel node importance evaluation method. First, the original clusterstructure of networks can be got by using the nontrivial eigenvectors, then, by using modularity to evaluate and merge these cluster structure, a cluster structure more fits forthe real networks can be got. Finally, the backbone graph can be extracted from the basenetworks by using cluster contraction principle, and then evaluating the importance ofgateway nodes in the backbone graph and using some super energy nodes to protectvital gateway nodes, this way can prolong the life of networks and improve thesurvivability of networks effectively.3. Most wireless sensor networks utilize static sink to collect data from themonitored region, such way may result in high traffic load in static sink’s vicinity. Thenodes located near static sink will be more requested than other nodes in networks, sothese nodes will consume more energy and trigger off route hole. Mobile sink has beendeveloped to solve route hole problem. However, latency and packet delivery delaycaused by mobile sink may be intolerable. A novel mobile data collector algorithm(MDCA) which deals with the route hole and the delivery delay well is proposed in thispaper. MDCA adopts the rule of packets intercept, by this rule, the intermediate nodecan intercept data packets coming from distant nodes that do not belong to propagationtree and forward these data packets through its bypass leading to the mobile sink nearby.This rule can reduce the transmitted distance of data packets efficiently, and improve thesecurity of data packets. In addition, in current designs, sensor node does not havecongestion self-adaptive function. Dealing with congestion in these reactive mannersmay result in longer delay and unnecessary packet loss. Hence, a self-adaptive rulebased on congestion monitoring model is also presented. The congestion node canrestore quickly by using bypass to split data traffic. This strategy performs better inputting down the congestion level and improving the success rate of packettransmission.4. A combination optimization routing algorithm (CORA) based on the dualchannel wireless sensor networks is presented to put down the blocking probability ofhigh load networks. This algorithm deals with date collision and multicast suppressionin channel competitive process well by using the dual-channel communication model.At the same time, this algorithm uses the infection sphere to reduce the number of nodeswhich join in researching the optimization route from the source node to the target node;this way can reduce energy consumption of networks. At last, this paper proposes acombination optimal routing algorithm with the help of a layered-graph model. Theblocked service in control plane can use the idle resource in data plane to transmit insynchronous manner, so the blocking probability of networks and the delay of communication can be cut down by this way. Simulation results show that thisalgorithm performs better in terms of the time consumption of communication and thetotal energy consumption. Compared with the other algorithms, the blocking probabilityof networks can be cut down13%.5. Aiming at measurement errors of distances between nodes in wireless sensornetworks, a distributed local robust minimum spanning tree algorithm (LRMST) whichis based on local minimum spanning tree algorithm and the robust discrete optimizationtheory is proposed in this paper. When uncertainties affect the objective parameter in themodel of minimum spanning tree, the problem of searching the minimum spanning treeof networks can be conversion to a discrete optimization problem by means of Booleprogramming model. For the objective parameter-uncertain Boole programming model,the theory of robust discrete optimization is proved that the robust counterpart of thismodel can be solved by disposing of one deterministic programming problem. So whenall the measurement distances between nodes have measuring errors unavoidably, thealgorithm proposed in this paper can obtain the robust minimum spanning tree bysolving one deterministic problem. The simulation results show that, with the changingof laboratory conditions, the sensor nodes in minimum spanning tree which was foundby LRMST algorithm have more higher degree, this character can ensure minimumspanning tree has remarkable survivability and robustness.更多还原