【作者】 柏荣刚；

【导师】 屈玉贵；

【作者基本信息】 中国科学技术大学 ， 通信与信息系统， 2009， 博士

【摘要】 无线传感器网络是一种特殊的无线通信网,因其节点数量巨大、成本低廉,可以快速部署,且不依赖于任何固定设施,能够实时、准确和全面地在多种场合下采集信息,从而改变人与自然的交互方式,被认为是21世纪最重要的技术之一。因为几乎所有节点都由电池供电,通常无法补充能源,所以节能是无线传感器网络设计中最重要的目标之一。而且从降低网络成本、节约资源和环境保护的角度来看,研究节能问题也具有重大的意义。本文在广泛调研的基础上,从MAC协议和跨层协议两方面,对无线传感器网络的节能协议进行了研究。并将人工智能理论应用于传感器网络的研究,提出了智能监测和覆盖算法。主要内容与创新点包括以下三个方面:1.提出了一种时分复用的MAC协议TDMA-WSN(Time Division MultipleAccess for Wireless Sensor Networks):首先,通过邻近节点间的广播,节点确定与其相冲突的节点集合。然后在时隙分配的过程中,为每个节点在时间帧中分配一个时隙,互相冲突的节点保证不会占用同一时隙。在分布式的传输数据过程中,因为节点只在自己的时隙内进行通信,所以避免了冲突。该协议不仅解决了隐藏节点问题,也避免了竞争、冲突和串音现象所带来的能耗,并证明了所提出的“冲突集合”是有效且最小的,使更多的节点可以同时无冲突地传输数据,提高了信道利用率。此外,分布式、自组织的信道访问模式,也减少了控制报文的开销。模拟实验表明,与现有的基于竞争和调度方式的MAC协议相比,TDMA-WSN协议能够有效地降低能耗,从而延长网络的生命期。2.提出了两个跨层协议CLWSN(Cross Layer Protocol for Wireless SensorNetworks)和TEPA(Energy-Efficient Tree-Structure Cross-Layer Protocol):这两个协议都是在MAC层和网络层之间交互信息,避免了传统协议追求节能而牺牲时延的缺点,提高了网络的综合性能。CLWSN协议融合了TDMA-WSN和树型路由协议。该协议对“冲突集合”的定义与TDMA-WSN不同,因为考虑了节点在路由树上的父子关系,从而减少了“冲突集合”的节点数量,提高了信道利用率。在分布式的时隙优化过程中,CLWSN协议进一步利用路由信息,重新为节点分配时隙,按顺序排列路径上的调度时间表,极大地降低了网络的传输延迟。TEPA协议也是在TDMA-WSN和树型路由协议间交互信息。与分布式的CLWSN协议不同,TEPA采用了收敛较快的微分进化算法,在基站集中式计算,为网络配置调度时间表。该算法将整个网络的调度时间表作为待进化的个体,在交叉和选择时参考了路由信息。其个体进化的目标方向是,尽量减少路由树上各路径间的冲突,加速单条路径上的数据传递,最小化网络的总延迟。TEPA协议首次折中考虑了在不同负载下网络的时延性能,计算出最优而非最短的MAC帧长度,从而提高了网络的吞吐量。模拟实验表明,CLWSN和TEPA协议在能耗与时延的综合性能方面有较明显的改善。3.提出了智能监测算法SOMSA(Self-Organizing Mapping MonitoringScheduling Algorithm)和智能覆盖算法SOMDA(Self-Organizing MappingDeployment Algorithm):本文将神经网络理论应用于无线传感器网络的节能研究,在SOMSA算法中,实现了传感器网络在动态变化环境中的智能监控。该算法使用自组织映射神经网络进行训练,节点不再孤立地感知信息,而是在簇内交换采集信息和探测模式,通过竞争选择最适应环境的节点,优胜者引领其它节点的进化方向,逐渐调整网络的探测模式,以适应所处的环境。训练结果表明,网络随着环境的不断变化,可以随时调整节点的采样频率,使得网络的探测模式与环境的变化规律一致,既提高了探测的精确度,又节省了感知单元的能耗。智能覆盖算法SOMDA,解决了传感器网络中对事件的自适应覆盖问题。首先,使用遗传算法将随机分布的网络组成二维网格,然后使用自组织映射算法对网络进行覆盖训练。在训练中,将发生的事件作为输入模式,将探测到事件的节点作为激活对象。被激活的节点修正自身位置,向事件的方向移动,随着时间的推移,网络逐渐趋于稳定。经过训练后,节点的分布精确地反映了事件的特征分布。模拟实验表明,SOMDA算法在不损失面积覆盖率的情况下,提高了探测事件的能力,均衡了节点的能耗。更多还原

【Abstract】 Wireless Sensor Networks(WSNs) are special wireless communication networks which are composed of a large number of nodes with low cost and simple structure. These nodes are easy to be deployed without any fixture.They can collect information quickly,accurately and comprehensively under a variety of occasions. WSNs make a strong impact on the intercommunion between human and the nature. And they are considered as one of the most important technologies in the 21st century.Since almost all nodes are supported by battery which are difficult to be supplied,energy efficiency is one of the most important goals in designing wireless sensor networks.Moreover energy-efficiency design is also significant for decreasing network maintenance cost and saving natural resources.This thesis investigates the hotspots of the research on the power-saving protocols for WSNs,studies the energy-efficiency issues on the MAC layer and the cross-layer protocols.Besides,it proposes a deployment algorithm and a monitoring algorithm based on the artificial intelligence theory.The main results and innovations of the thesis are as follows:1) Proposes a TDMA MAC protocol TDMA-WSN:First of all,the nodes broadcast between the neighborhoods to find the conflicting sets.Then each node is allocated a slot in the frame in the slot allocation process.The nodes which are conflicting with each other will not be assigned the same slot.Then the nodes could transmit data without collision during their own slots in the data transmission process.TDMA-WSN can solve the hidden nodes,collision, competition and overhearing phenomenon,and prove that the conflicting set is effective and minimal which increases the number of nodes sending data at the same time without collision and improves the throughput.Moreover,the self-organized and distributed channel access pattern decreases the mount of control packets.Simulation results indicate that TDMA-WSN outperforms the traditional contention-based and scheduling-based protocols in terms of energy consumption and prolongs the lifetime of the networks.2)Proposes two Cross-layer protocols CLWSN and TEPA:CLWSN and TEPA both exchange information directly between the network layer and the MAC layer,which can avoid the disadvantage of large delay in the traditional protocols.They optimize energy utilization and improve the integrated performance of the network.CLWSN fuses the TDMA-WSN and the tree routing protocol.However,the definition of the conflicting set is different from that in the TDMA-WSN.It reduces the number of the nodes in the conflicting set to increase the throughput by using the routing information from the tree.Besides,in the slot optimization scheme,it optimizes the slot assignment in the path of the tree by rearranging the order of the slots to avoid the disadvantage of large delay with the routing information.TEPA also fuses the TDMA-WSN and the tree routing protocol.On one hand, TEPA is different from the distributed CLWSN that the BS calculates centrally to assign the TDMA schedule for the whole networks using the Differential Evolution (DE) algorithm which has fast convergence speed.In DE,the candidate is the schedule of the networks,and the routing informations are referenced in the crossing and selecting operations.To minimize the total delay of the networks,the direction of the evolution is limiting contention between adjacent branches and accelerating transmission in individual branch of the data gathering tree.On the other hand,it is different from the previous schemes that TEPA does not try to minimize the frame length but choose a suitable schedule length based on the tradeoff on throughput and delay under different traffic load.Simulation results show that CLWSN and TEPA enhance the integrated performance in terms of delay and energy consumption.3)Proposes the intelligent monitoring algorithm SOMSA and the intelligent deployment algorithm SOMDAThis thesis studies the energy saving algorithms in WSNs using neural networks theory.SOMSA implements the intelligent monitoring in the variable environment, which is based on artificial neural-networks self-organizing maps(SOM) algorithm. During the training process,the nodes do not work alone but compete and cooperate to choose the victors which are the most adaptive to the environment.Then the victors guide the evolution direction,and adjust their neighbors’ internal prototype vectors and the sampling frequency to study the input patterns of the environment.At the end of the training,the sampling frequency of the network will reflect the variety of the environment to save the energy consumption of the sensing module and increase perceptive precision.Intelligent deployment algorithm SOMDA resolves the self-adaptive deployment problem in the sensor networks.First of all,the randomly distributed nodes need to be arranged in a two-dimensional lattice.During the SOM training process,the input patterns are the event location.The nodes which detect the events are active,and the activated nodes update their location following the events.After going through the training,the networks keep stable and the statistical distribution of the nodes approaches that of the events in the interest area.The simulation results indicate that SOMDA does not lose the coverage rate but enhance the performance in terms of detecting ability and energy equalization.更多还原