【作者】 王艳；

【导师】 池小波；

【作者基本信息】 山西大学 ， 计算数学， 2019， 硕士

【摘要】 近年来,多传感器信息融合估计受到了越来越多的关注,并广泛应用于移动机器人,智能电网等领域,基于无线传感器网络(WSNs)的分布式融合估计问题已经成为国内外的研究热点.例如,移动机器人追踪,多传感器三容水箱液面估计以及温室区域温度控制估计等实际问题都是基于分布式估计或融合估计理论的.然而基于WSNs的融合估计框架中,由于网络的介入使得数据在传输过程中会受到时延和丢包等网络诱导特征以及恶意的网络攻击等影响,这些因素都可能导致估计性能下降,也使得融合估计系统的分析和设计面临诸多挑战.因此如何降低多传感器信息传输过程中诱导的网络不确定因素对状态估计的影响,是基于WSNs的分布式估计研究的难点之一,设计一个鲁棒性强,可靠度高的分布式滤波算法是实现信息融合问题的关键所在,但是针对这些问题的研究工作还较少.考虑到多传感器多率异步采样机制,传输过程中时延和丢包等网络诱导特征以及网络攻击,本文研究基于WSNs的分布式融合估计问题,研究的主要工作及取得的创新性成果如下(1)综述了多传感器分布式融合估计的国内外研究现状和存在的挑战性问题.(2)研究带有模型不确定性,网络诱导时延和数据丢包的无线传感网络化系统分布式融合估计问题,设计一种基于鲁棒Kalman滤波分布式融合估计器.针对不同传感器在数据传输过程中出现的多通道网络诱导时延,利用补偿机制将时滞系统转化为无时延系统进行研究,并利用多个伯努利序列来分别建模不同通道的数据丢包现象.基于成功到达的采样数据,设计一组基于鲁棒Kalman滤波算法的网络化估计器来得到局部估计,并利用协方差交互融合方法来减少通信和计算负担,获得比局部估计精度更高的融合估计器.(3)进一步研究一类带有多传感器多率采样的网络化系统分层分簇融合估计问题.考虑到传感器节点的空间分布性和能量受限等特征,提出一种基于簇内局部估计和簇间融合估计的双层异步分布式估计策略.考虑无线传感器网络的丢包和网络攻击现象,在簇头节点设计一个基于簇内节点多率采样数据的网络化强跟踪滤波器来获得系统的局部估计,其中引入的时变渐消因子可以补偿由异步采样和未知系统干扰导致的建模误差.其次,利用多个邻居簇头节点的局部状态估计或预测状态估计信号,提出一种带有时变融合时刻的多率融合估计方法.所提方法充分考虑了多传感器多率采样方式、多率数据的多通道传输以及分层分簇的分布式结构,适用于异步多传感器分布式估计系统.本文针对带有多传感器多率异步采样,传输过程中时延、丢包和网络攻击等网络诱导特征分布式融合估计问题进行研究,设计了两类滤波融合方法,分别对移动目标追踪和三容水箱ITTS进行数值仿真,其实验结果验证了所提分布式融合估计方法的可行性.更多还原

【Abstract】 In recent years,multi-sensor information fusion estimation has attracted more and more attention and has application in a broad range of areas such as mobile robots,smart power grids and so on.Distributed fusion estimation based on wireless sensor networks(WSNs)has become a research hotspot at home and abroad.For example,practical problems such as WSNs-based mobile robot tracking,multi-sensor three-tank liquid level estimation and greenhouse temperature control estimation are based on distributed estimation or fusion estimation theory.However,in the fusion estimation framework,data will be affected by network-induced characteristics such as delay and packet loss and malicious network attacks during date transmission process,these factors may lead to the degradation of the estimation performance,and also make the analysis and design of the fusion estimation system face many challenges.Therefore,how to reduce the influence of network uncertainty factors induced in the process of multi-sensor information transmission on state estimation is one of the difficulties in distributed estimation research based on WSNs.The key to information fusion is to design a highly robust and reliable distributed filtering algorithm.but there is less research work on these issues.Considering multi-sensor multi-rate asynchronous sampling mechanism,network induced characteristics such as delay and packet loss during transmission and cyber-attack,this paper studies WSN-based distributed fusion estimation.The main work and innovative achievements are as follows(1)This paper reviews the research status and challenges of distributed multi-sensor fusion estimation at home and abroad.(2)The distributed fusion estimation problem of wireless sensor network system with model uncertainty,network-induced delay and data packet loss is studied,and a distributed fusion estimator based on robust Kalman filter is designed.In view of the time-delay induced by multi-channel network in the process of data transmission of different sensors,the time-delay system is transformed into a delay-free system by using the compensation mechanism,and multiple Bernoulli sequences are used to model the data packet loss phenomenon of different channels.Based on the successfully arrived sampling data,a set of networked estimators based on robust Kalman filtering algorithm is designed to obtain local estimators,and the covariance interaction fusion method is used to reduce the burden of communication and calculation,so as to obtain a fusion estimator with higher accuracy than local estimation.(3)This paper further studies the hierarchical clustering fusion estimation problem for a class of networked systems with multi-sensor and multi-rate sampling.Considering the characteristics of spatial distribution and energy limitation of sensor nodes,a twolayer asynchronous distributed estimation strategy based on intra-cluster local estimation and inter-cluster fusion estimation is proposed.Considering packet loss and cyber-attack of wireless sensor network,a networked strong tracking filter based on multi-rate sampling data of nodes in the cluster is designed in the cluster-head node to obtain the local estimation of the system,the time-varying fading factor can compensate the modeling errors caused by asynchronous sampling and unknown system interference.Furthermore,a multirate fusion estimation method with time-varying fusion time is proposed by using local state estimation or predicted state estimation signals of several neighbor cluster head nodes.The proposed method takes full account of multi-rate sampling,multi-rate data transmission and distributed structure of hierarchical clustering,and is suitable for asynchronous multisensor distributed estimation system.In this paper,the distributed fusion estimation problem of networks-induced characteristics such as delay,packet loss and cyber-attack during the transmission process with multi-sensor and multi-rate asynchronous sampling is studied.Two kinds of filtering fusion methods are designed to respectively carry out numerical simulation on moving target tracking and three-tank ITTS.The simulation results verify the feasibility of the proposed distributed fusion estimation method.更多还原