一类不确定离散复杂动态网络状态估计

发布时间：2018-07-06 15:44

本文选题：复杂动态网络 + 单重丢包　；参考：《南京邮电大学》2017年硕士论文

【摘要】：以往复杂网络的同步控制、拓扑辨识、故障诊断等领域的研究,往往都是假设在可获得网络节点状态的前提下进行,但是实际的网络节点规模庞大、拓扑结构复杂,一般只能通过测量得到网络节点的部分状态信息。因此,为了进一步研究与分析复杂网络的动力学特性、拓扑辨识以及故障诊断,构造设计复杂网络状态估计器估计原网络节点状态具有非常重要的实际意义。在实际网络中,由于传感器本身的物理特性和技术的限制以及网络拥塞、网络故障、网络遭受攻击等原因,在测量网络信号时需考虑传感器随机饱和现象以及数据丢包现象。同时由于网络规模的不断增长,信道带宽资源往往受到限制。因此,综合复杂网络存在以上实际因素,我们深入研究一类不确定离散复杂动态网络状态估计器的设计问题。本文主要内容和所取得的研究成果如下:1.针对一类同时出现随机丢包和传感器饱和现象的离散复杂动态网络,对其状态估计问题进行了研究。网络的丢包现象利用满足Bernoulli分布的随机变量进行描述,传感器的饱和现象利用饱和函数进行刻画。通过稳定性理论并结合随机分析的方法,以线性矩阵不等式形式,给出上述状态估计器增益设计准则。最后,实验仿真验证本文所提方法的有效性。2.针对一类存在随机丢包的离散复杂动态网络,研究其基于事件触发的状态估计方法。利用满足Bernoulli分布的随机变量描述网络的丢包现象,所设计的事件触发估计器可以减少通信信道冗余数据的传输,节省网络带宽资源,提高其利用率。结合稳定性判据和随机分析的方法,以线性矩阵不等式形式,给出上述状态估计器增益设计准则。通过实验仿真,验证本文所设计的状态估计器可以有效地估计出原网络节点状态。3.针对一类存在随机拓扑跳变和丢包的输出耦合离散复杂动态网络,设计异步状态估计器。考虑复杂网络内外耦合拓扑矩阵满足马尔可夫跳变,利用满足Bernoulli分布的随机变量描述网络的丢包现象,所设计的异步观测器克服了原网络拓扑结构未知或存在随机拓扑跳变的问题。基于稳定性理论和随机分析的方法,以线性矩阵不等式形式,给出上述状态估计器增益设计准则。通过实验仿真,验证本文所提方案的有效性。
[Abstract]:In the past, the research of synchronous control, topology identification, fault diagnosis and other fields of complex networks is often carried out on the assumption that the network node states can be obtained, but the actual network nodes are large in scale and complex in topology structure. Generally, only part of the network node state information can be obtained by measurement. Therefore, in order to further study and analyze the dynamic characteristics, topology identification and fault diagnosis of complex networks, it is of great practical significance to construct and design a complex network state estimator to estimate the node state of the original network. In practical networks, due to the physical characteristics and technical limitations of sensors, network congestion, network failures, network attacks and other reasons, sensor random saturation and data packet loss should be considered in the measurement of network signals. At the same time, due to the continuous growth of network size, channel bandwidth resources are often limited. Therefore, the design of a class of uncertain discrete complex dynamic network state estimators is studied. The main contents of this paper and the research results obtained are as follows: 1. The state estimation problem of a class of discrete complex dynamic networks with simultaneous random packet loss and sensor saturation is studied. The packet loss phenomenon of the network is described by random variables satisfying Bernoulli distribution, and the saturation phenomenon of the sensor is characterized by the saturation function. Based on the stability theory and stochastic analysis, the gain design criteria of the state estimator mentioned above are given in the form of linear matrix inequality (LMI). Finally, the experimental results show that the proposed method is effective. 2. A state estimation method based on event triggering for a class of discrete complex dynamic networks with random packet loss is studied. Using random variables satisfying Bernoulli distribution to describe the packet loss in the network, the event trigger estimator is designed to reduce the transmission of redundant data in the communication channel, save the bandwidth resources of the network and improve its utilization ratio. Combined with the stability criterion and the stochastic analysis method, the gain design criteria of the state estimator are given in the form of linear matrix inequality (LMI). The experimental results show that the proposed state estimator can effectively estimate the state of the original network node. An asynchronous state estimator is designed for a class of output coupled discrete complex dynamic networks with random topology hopping and packet loss. Considering that the topological matrix inside and outside the complex network satisfies the Markov jump, the random variables satisfying Bernoulli distribution are used to describe the packet loss phenomenon of the network. The designed asynchronous observer overcomes the problem that the topology of the original network is unknown or there is a random topology jump. Based on the stability theory and stochastic analysis method, the gain design criteria of the state estimator mentioned above are given in the form of linear matrix inequality (LMI). The effectiveness of the proposed scheme is verified by experimental simulation.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5

【相似文献】