面向服务的无线传感器网络关联感知策略与容错协同机制研究

The critical bottleneck that has limited the future development of wireless sensor networks is how to support complicated and practical applications with simple and low-cost sensor nodes. Current research works had not fully considered the special style of wireless sensor networks, such as strong connectivity, poor reliability and spatio-temporal correlationship, and they generally focused on some local objectives while efficent global solutions were rather scare. This proposal is dedicated to consider the service-oriented trend together with the special characteristics of the wireless sensor network. We propose to model the spatial-temporal correlationship and probability distribution among the sensed data with the Markov random field, and then design the node selection and sensing strategy with data accuracy guarantee as well as the implementation mechanism which can efficiently hide the resource different; We propose to develop a new processing model by combining the data aggregation and compression methods together so that data information can be re-constructed with nonlinear optimization, and propose to build reliable transmission mechanism for both data and control information with fault tolerance in different layers, such as access control, path selection and topology conectivity, and the mechanism efficiency is also evaluated by building the multi-status Markov chain and these parameters are also optimized in depth by measuring their importance in the model;Further,we propose to study the periodic scheduling problem in the multi-service wireless sensor networks, and build low-complexity mechanism and methods to provide different services by using dimensionality reduction and divide-and-rule strategies, and aim at solving the problems of inefficiency and waste of resources originated from the traditional protocols and providing highly reliable service in heterogeneous and unstable environments.

如何通过简约的传感节点来支持复杂的实际应用，已成为束缚无线传感器网络未来发展的关键瓶颈。现有研究成果未能较好结合其强连通度、低可靠性及时空关联等独特的网络形态，只针对某些局部目标而展开，缺乏有效的全局解决方案。本项目将面向服务的发展趋势和网络自身特性相结合，提出利用马尔可夫随机场对感知数据的时空相关性和概率分布情况进行建模，设计带质量保证的节点选取和感知策略及可有效隐藏资源差异性的实现机制；提出将融合和压缩相结合的新处理模型以实现信息重构的非线性优化，并通过接入控制、路径选取和拓扑连接等多层次的容错协同来建立数据和控制信息的可靠传输机制，构造多状态马尔可夫链评估其效率，分析各度量参数并进行深层次优化；进一步研究多服务背景下网络周期调度问题，提出基于降维和分而治之等策略建立具备低复杂度的差异化服务的机制和方法，通过跨层协同来解决传统方法存在的问题，实现在异构不稳定环境下提供高可靠服务的目标。

“构建惠及全民、以用户为中心、无所不在的信息网络体系”这一目标体现了我国全面进入信息时代所面临的重大战略需求，而无线传感器网络是实现该目标最具前景的有效手段。围绕着如何通过简约的传感节点来支持复杂的实际应用这一瓶颈问题，本课题结合无线传感器网络强连通度、低可靠性和时空关联等特点，从网络体系不同层次出发，研究并建立以服务为中心、以关联感知和容错协同为主要技术手段的理论模型和实现机制，以期达到在异构不稳定环境下提供高可靠服务的目标。.课题组在以下几个方面取得重要成果：1）在节点感知调度策略问题上，提出了一种基于平衡覆盖集的关联节点集计算方法和选取调度策略，提出采用马尔可夫随机场模型（MRF）对感知数据间的关联性进行建模的新方法，并在此基础上解决了无线传感器网络中带误差保证的单服务和多服务情况下的节点调度问题。实验结果表明，所提出的算法明显优于目前国际上的主流方法；2）在丢失数据修复问题上，研究并分析数据丢失的几种模式，在此基础上利用时空相关性，提出了基于曲面拟合和基于马尔科夫随机场的数据重建方法，与传统方法相比，所提出算法具有稳定的效果；3）在无线传感器网络数据质量控制问题上，研究在数据清洗过程中不同指标间的关系对质量评估的影响，通过对比分析多种不同顺序的数据清洗方案，提出了基于数据质量指标间关系的无线传感器网络数据清洗策略。实验表明，该数据清洗策略拥有较好的清洗效果；4）在数据备份问题上，建立时空冗余数据清除模型，提出了基于簇的数据备份策略，同时采用健壮节点来备份普通节点的感知数据，实验结果表明，在有效提升网络的寿命的同时，所采用的备份算法具备良好的容错性。.本项目深入研究了无线传感器网络所面临的突出问题和主要矛盾，对于本领域其它关联问题的解决，以及推动无线传感器网络理论研究和实际应用部署都具有较高的科学价值；这些研究还将为未来无线网络和互联网技术提供较新的研究方法和研究思路。