具有通信约束和信息不完全的网络化系统的控制和滤波

发布时间：2018-05-11 11:37

本文选题：随机多步时延 + 多包丢失　；参考：《华东理工大学》2015年博士论文

【摘要】：随着信息技术的不断发展,计算机网络已经成为现代通信技术的核心,被应用在生活中的方方面面。网络化控制系统则将网络的概念引入到了控制系统中,利用网络来连接控制系统中的各个元部件。由于网络化控制系统具有诸多优点,在近年来得到了迅猛发展,逐渐代替了传统点对点控制系统,被越来越多地应用于实际系统中。然而,网络的引入同时也带来了许多挑战。网络诱导时延,数据包丢失以及量化误差是由于网络带宽限制,传输通道拥堵等原因所引起的最常见的三种网络化系统特性。本文针对这三种网络特性,对网络化控制系统展开了研究。论文分析了随机多步网络诱导时延,随机连续多数据包丢失以及量化误差在网络化控制系统中的多发性,构建了新的数学模型来分别描述这三种特性。本文以解决具有该类通信约束和信息不完全的网络化系统的控制和滤波问题为研究目标,分别设计了无偏最小协方差滤波,H∞滤波,集员滤波,H∞控制,量化H∞控制等方法,并对所得结果进行了仿真验证。主要内容包括以下几个部分： (1)随机多步传感器时延在实际网络化控制系统中的发生概率远远高于随机单步传感器时延或确定性时延。根据这一特性,提出描述随机多步传感器时延的新模型,并基于该模型,为具有随机多步传感器时延的网络化系统设计了无偏最小误差协方差滤波方法。提出滤波器无偏条件以及相应的滤波算法,并给出了设计和证明过程。通过MATLAB仿真实验,验证了提出的无偏最小误差协方差滤波方法的有效性和优越性。 (2)根据网络化系统中随机连续多数据包丢失现象的常发性,建立了可以有效描述最大丢包数已知的随机连续多数据包丢失现象的新模型。通过该数学模型,设计了改进的H∞滤波方法,并给出了相应的滤波算法。通过仿真实验验证了该方法能够有效地解决具有随机连续多数据包丢失的网络化系统问题。 (3)在实际系统中,噪声信号往往是未知有界的。根据这一特性,分析了集员滤波方法在解决未知有界噪声问题时的优势。通过对数量化器来描述系统的量化特性,为一类既具有未知有界噪声,又具有量化误差的网络化控制系统设计了集员滤波器。根据滤波器的设计过程,给出相应的集员滤波算法,并通过仿真实验验证了该滤波方法的有效性。 (4)根据随机多步传感器时延模型,建立了随机多步传输时延模型。提出具有随机多步传输时延的网络化控制系统的量化H∞控制问题。设计了量化H∞控制策略,使得闭环控制系统渐近均方稳定,并且控制输出满足H∞性能指标。根据设计过程,提出了量化H∞控制算法并通过仿真实验验证其有效性。 (5)分析线性变参数框架在处理非线性系统问题中的优越性,提出了一类新型的具有随机多步传感器时延的网络化线性变参数系统的控制问题。设计了改进的H∞控制器,并通过扩展CCLM算法,求得了控制器参数,使得闭环系统渐近均方稳定,并且控制输出满足改进的H∞性能指标。最后,通过仿真实验验证算法有效性。
[Abstract]:With the continuous development of information technology, the computer network has become the core of modern communication technology and is applied to all aspects of life. The networked control system introduces the concept of network into the control system and uses the network to connect the elements in the control system. Because the networked control system has many advantages, In recent years, the traditional point to point control system has been gradually replaced by the traditional point to point control system, which has been used more and more in the actual system. However, the introduction of network has also brought many challenges. Network induced delay, data packet loss and quantization error are the most common causes due to network bandwidth constraints, traffic congestion and so on. In view of the characteristics of the three networked systems, this paper studies the networked control systems in view of the three network characteristics.
In this paper, a new mathematical model is constructed to describe the three characteristics of the random multistep network induced delay, random continuous multiple data packet loss and quantization error in the networked control system. In this paper, the control and filtering problems of networked systems with such communication constraints and incomplete information are solved in this paper. The objective is to design unbiased minimum covariance filter, H filter, collector filter, H infinity control, quantized H infinity control and so on. The results are simulated and verified. The main contents include the following parts:
(1) the occurrence probability of random multistep sensor delay in the actual networked control system is far higher than that of random single step sensor delay or deterministic delay. Based on this characteristic, a new model describing the delay of random multistep sensor is proposed. Based on this model, the unbiased optimum is designed for a networked system with random multistep sensor delay. The method of small error covariance filtering is proposed. The unbiased condition of the filter and the corresponding filtering algorithm are proposed, and the design and proof process are given. The effectiveness and superiority of the proposed method of unbiased minimum error covariance filtering is verified by the MATLAB simulation experiment.
(2) according to the frequent occurrence of random continuous multiple packet dropout in the networked system, a new model of random continuous multiple packet dropout which can effectively describe the maximum packet loss number is established. Through this mathematical model, an improved H filtering method is designed and the corresponding filtering algorithm is given. The method can effectively solve the problem of networked systems with random continuous multiple packet dropout.
(3) in the actual system, the noise signal is often unknown and bounded. According to this characteristic, the advantage of the set member filtering method in solving the unknown bounded noise problem is analyzed. The quantitative characteristics of the system are described by the logarithm quantizer, and a set of networked control systems with unknown bounded noise and quantized error is designed. According to the design process of the filter, the corresponding set membership filtering algorithm is given, and the validity of the filtering method is verified by simulation experiments.
(4) based on the random multistep sensor delay model, a random multistep transmission delay model is established. The quantization H infinity control problem for networked control systems with random multistep transmission delay is proposed. A quantized H infinity control strategy is designed to make the closed-loop control system asymptotically mean square and the control output satisfies the performance index of H infinity. A quantization H infinity control algorithm is proposed and its effectiveness is verified by simulation experiments.
(5) the superiority of the linear variable parameter frame in dealing with the nonlinear system problem is analyzed. A new type of networked linear variable parameter system with random multistep sensor delay is proposed. An improved H controller is designed. The controller parameters are obtained by extending the CCLM algorithm, which makes the closed-loop system asymptotically stable and square stability. And the control output satisfies the improved H infinity performance index. Finally, the effectiveness of the algorithm is verified by simulation experiments.

【学位授予单位】：华东理工大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TP273

【参考文献】