【作者】 阎毓杰；

【导师】 王殊；

【作者基本信息】 华中科技大学 ， 信息与通信工程， 2007， 博士

【摘要】 无线传感器网络是由部署在监测区域内大量的微型传感器节点通过无线电通信形成的一个多跳的自组织网络系统,其目的是协作地感知、采集和处理网络覆盖区域里监测对象的信息,并发送给观察者。由于网络自身的动态性以及所处环境变化的不可预测性,无线传感器网络需要具备自组织管理的能力,即具备可重构的功能,在任何时刻任何地方能够快速展开并自动组网,自行进行配置和维护管理,合理高效地协调网内资源与任务的分配调度,自动分发和收集传感数据。移动代理技术因其转移计算代码、交互本地化的特征给分布式系统的研究带来了新的启发。本论文根据无线传感器网络的特点,将移动代理引入进来,借助其自主迁移和分布式计算能力,并结合微观经济学以及博弈论的观点与方法,以解决无线传感器网络自组织管理中资源与任务的互优化以及数据流的分发与收集等关键技术。本论文所开展的研究工作主要有以下这些:在移动代理的模型设计方面,本论文主要将基于移动代理的计算模型和基于C/S的计算模型的性能进行理论分析和比较;在设计移动代理自身结构的基础上将移动代理技术中间件化,构建基于移动代理的无线传感器网络的软件架构,以形成移动代理中间件的应用框架;还探讨了移动代理的路由问题,并提出了移动代理的静态/动态路由策略。在自组织资源与任务管理方面,本论文探讨移动代理模式下的资源与任务互优化技术。主要运用微观经济学中的市场机制原理,在移动代理的模式下分别将节点和任务智能化,设计了自组织的微观经济模型框架;并在此基础上提出了基于市场机制的资源分配策略MRA和任务调度策略MTS,以在有效调度执行网络任务的同时合理优化分配网内资源,实现网内资源与任务的互优化。在自组织数据流管理方面,本论文探讨移动代理模式下如何通过促进多节点协作来实现数据流的有效分发与收集。关于数据流收集的研究,主要提出了一种基于分辨率的并行量化交叠的数据融合算法RPQO,它采用多移动代理并行执行融合任务,同时使用量化编码的办法减少融合过程中的数据量。关于数据流分发的研究,主要是结合博弈论的方法,在节点的转发能耗和吞吐量之间寻求折中,提出了基于帕累托最优效用的包转发算法POUPF和子博弈精炼均衡的包转发策略。本论文的研究成果一方面可以用来支撑无线传感器网络其它关键技术(如路由、中间件技术等)的研究;另一方面也可以运用于实时性强、对网络自组织性要求高的应用场景(如目标追踪)。更多还原

【Abstract】 Wireless sensor networks(WSNs) is a distributed technology for sensing and collecting environment information, consisting of large numbers of small, low-powered, wireless“motes”each with limited computation, sensing, and communication ability. However, since there are lots of unpredictable conditions in WSNs and its environment, WSNs must possess the perfect capability of self-organizing management so that the network can reconstruct or reconfigure itself automatically, harmonize the resource allocation and task scheduling effectively, and transmit data flow by using apporiate protocols or algorithms.With the trait of moving computation to data, the mobile agent technology brings a new idea for the distributed system design. Therefore, by introducing the mobile agent into WSNs and absorbing the ideas of microeconomics and game theory, a deep and systematic research work is done in this thesis to design the mobile agent-based self-organizing management technologies with focus on the mobile agent model, the optimization of resource & task, the transmiting and collecting of data flow.For the mobile agent model, this thesis analyses the mobile agent-based computing model and compares it with the client/sever-based computing model, designs the mobile agent middleware which supports the application framework of WSNs with the mobile agent’s architecture. For the problem of mobile agent route, a static routing method is design to search an optimal route with genetic algorithm, and the dynamic route method is also discussed.For the self-organizing management on resource and task, this thesis designs a self-organizing microeconomic system in which the application tasks are scheduled onto nodes by mobile agents while distribute their resource consumption across network. Further, it proposes the market-based resource allocation policy named MRA which satisfies the optimal division of the single capacity for multiple tasks. For harmonizing the resource allocation and task scheduling, it also proposes a market-based task scheduling policy named MTS which schedules tasks to the set of optimal nodes.For the self-organizing management on data flow, this thesis provides a cooperative solution for data transmiting and collecting. For data collecting, it proposes a resolution based parallel quantizing overlapping algorithm named RPQO to execute the distributed data fusion. For data transmiting, it proposes a pareto optimal utility based packet forwarding algorithm named POUPF which constitutes a Nash Equilibrium. Also, a subgame perfect method is achieved in the repeated game of packet forwarding.The achievements of the research work in this thesis can be used to support other key technologies of WSNs. And it can also be used in the applications with the requirements of real-time and dynamic adaptation, such as target tracking.更多还原