【作者】 姜翀；

【导师】 张庆灵；

【作者基本信息】 东北大学 ， 控制理论与控制工程， 2010， 博士

【摘要】 网络控制系统是通过实时网络构成闭环的反馈控制系统。在网络控制系统中,传感器、控制器、执行器及被控对象和通讯网络连接在一起,设备之间信息通过一个共享的网络通道进行传输。网络控制系统是计算机技术、通信技术与控制技术发展与融合的产物,与传统的“点对点”直接相连的控制系统相比,网络控制系统具有布线简单、结构灵活、易于安装与维护、有较高的容错与故障诊断能力等优点。然而,由于网络带宽有限以及信息传输过程中的碰撞,不可避免地导致信息延迟、数据包丢失、多包传输以及时变采样周期等问题。这些问题的存在破坏了系统性能,甚至导致系统不稳定。因此,如何针对网络控制系统自身特点,应用已有的控制理论方法来解决网络控制系统中存在的问题,已成为控制领域研究的一个热点。目前,针对具有网络诱导时延及数据包丢失网络控制系统的建模、稳定性分析及控制器设计等问题的研究取得了一系列的成果。而关于如何有效使用有限通讯资源,使数据信息合理分享通讯渠道,从而提高网络控制系统性能的问题尚未得到充分研究。本文在总结前人工作的基础上,针对具有网络诱导时延和数据包丢失的网络控制系统,基于Markov跳变系统的相关理论,提出了具有随机时延及数据包丢失的网络控制系统的建模及确保闭环系统稳定的静态输出反馈H_∞控制器的设计方法;提出了经由周期时变局部控制器镇定的网络控制系统的建模方法,使控制信息有效利用带宽有限的通讯渠道,从而提高系统性能;提出了具有量化反馈的网络控制系统的建模方法,通过采用动态量化器,调节其动态调节参数,缩小量化范围,使得系统渐近稳定且具有良好的H_∞性能。本文的主要工作包括以下几个方面:（1）针对网络控制系统中网络带宽有限的特点,为了节约网络带宽,降低网络拥塞率,缩短数据传输时延,我们根据Markov跳变系统的相关理论,提出了一类具有随机时延以及数据包丢失的网络控制系统的建模方法。其中,系统中的传感器、控制器采用时间驱动,执行器采用事件驱动,传感器的数据采用单包传输。在实际控制中,因为被控对象的状态无法被直接测量得到,所以提出了一种在实际应用中更容易实现的静态输出反馈H_∞控制器设计方法。根据线性矩阵不等式相关理论,通过建立和求解一个凸优化问题,得出了静态输出反馈H_∞最优控制律。与已有的静态输出反馈控制器的设计方法相比,我们的设计更简单可行。（2）研究在介质访问受限的条件下网络控制系统的镇定问题。根据周期时变系统理论,研究了由网络连接具有多个传感器、控制器和执行器存在的网络控制系统的建模问题。由于介质的传输渠道带宽是有限的,在一定的时间内只有部分的传感器和执行器可以获得由控制器输送的变换信息,而其他的传感器和执行器只能等待,因此网络控制系统的性能不仅依赖于控制器的选择,还依赖于介质在网络控制系统中的传输效率。我们通过设计彼此经由网络连接的分散动态输出反馈控制器,得出网络控制系统镇定的充要条件。根据线性矩阵不等式相关理论,给出一整套周期时变局部控制器的设计方法。（3）研究了具有网络诱导时延及数据包丢失的网络控制系统新模型,模型中考虑了数据信息通过网络传输时特有的采样及量化等特征。分别讨论了利用量化状态反馈实现网络控制系统H_∞控制及具有非线性扰动网络控制系统的镇定问题。基于线性矩阵不等式相关理论,推导出确保系统渐近稳定同时满足H_∞性能的状态反馈控制器的设计方法。通过求解凸优化问题,获得网络控制系统及具有非线性扰动的网络控制系统的最大允许时滞界。（4）研究了一个新的网络控制系统模型,此模型同时考虑了数据信息在传感器到控制器及控制器到执行器之间传输过程中存在网络诱导时延及数据包丢失等情况,以及多率采样及动态量化作用对系统的稳定性所产生的影响。通过选取适当的Lyapunov-Krasovskii函数,得到一个不仅依赖于系统的控制输入信号,还依赖于系统的测量输出信号量化动态输出反馈控制器。根据每个采样时刻所确定的动态调节参数μ_i（i=1,2）的值（μ_i（i=1,2）取自可数集且依赖于系统的控制输入和测量输出值）设计量化控制器,且控制器设计仅仅需要量化器q_μ1（y_p（t））和q_μ2（y_c（t））在采样序列t=s_i,t=σ_i（i=1,…,∞）的值。为了获得较少保守性镇定策略,文中引入一些松弛变量,利用线性矩阵不等式方法得出不确定时变网络控制系统的最大允许时滞界。与已有的结论相比,我们的结论具有较少的保守性。（5）研究了一类不确定的网络控制系统新模型。此模型考虑了数据信息在传感器到滤波器之间传输过程中存在的网络诱导时延及数据包丢失等情况,同时考虑了采样及量化作用对系统的影响。通过选取适当的Lyapunov-Krasovskii函数,得出H_∞滤波器的设计方法。在设计中,动态量化控制器设计依赖于每个采样时刻所确定的动态调节参数μ的值（μ取自可数集且依赖于系统的测量输出值）,且控制器设计仅仅需要量化器q_μ（y_p（t））在采样序列t=s_i（i=1,…,∞）的值。然后,研究了更具有一般性的多面体不确定网络控制系统的滤波器设计问题。通过建立求解一个凸优化问题,提出一个使系统具有更好的H_∞性能的滤波器设计方法。更多还原

【Abstract】 Networked control systems（NCSs） are closed loop feedback systems constituted by real-time network.In NCSs,sensors,actuators and controllers are connected by means of a network or other shared medium,and information transmission among control system components is realized through a commonly shared network medium, which is the development and integration of control science,computer technology and network communication.Compared with the control system marked by traditional "point to point" direct link,there are several advantages for NCSs,which are given as follows,but not limited to less wire,less cost of installation,easy maintenance, higher flexibility and reliability,good fault detection ability,etc.However, some problems such as network-induced delay,packet dropout,multi-packet transmission and time-varying sampling period are usually arisen due to band-limited channels,so the use of a network may deteriorate the performance or cause instability. Consequently,associating the characteristic of NCSs,it has become a hot issue how to solve these problems in NCSs by utilizing existing control theories.Recently,a wide range of research has been reported dealing with problems related to modeling,stability analysis and design of controller of NCSs with network-induced delay and packet dropout.However,the problem how to efficiently utilize the finite communication resources and sufficiently share network bandwidth is still to be solved,which can guarantee stability of closed-loop system and maintain related good performance.On the basis of integration of related previous works, model formulation for NCSs with stochastic time delay and data dropout is proposed by utilizing the theory of Markov jump system,and static output feedback controller is designed in this dissertation.Furthermore,the model formulation for NCSs is proposed by virtue of periodically time-varying controller,which facilitates the share of information with the limited communication channel.Hence,the performance of system is improved.By using the dynamical scalar,the dynamical parameters can be switched,and the proposition of formulation of NCSs with quantized feedback controller can stabilize the model system and maintain good H_∞function.The main results obtained in this dissertation are summarized as follows:（1）Considering the limitation of bandwidth in the NCSs,a class of formulation methods for NCSs with stochastic delay and data packet dropout are proposed in terms of the theory of Markov jump systems,which is beneficial to saving the bandwidth,reducing congestion of network,shortening delay of data transmission. In the designed models,sensor node and controller node are time-driven,the actuator node is even-driven,and the sampled data were lumped together into single data packet.Since the state of NCSs can’t be measured directly,the H_∞controller design is presented through static output feedback.By means of the LMI,the design of the H_∞optimal controller is formulated as a convex optimization problem.Compared with the existing results,the merits of the obtained results are simple and feasible.（2） The stabilization problem of NCSs is studied under conditions of restricted media access NCSs.Based on the theory of time-varying periodic system,the modeling of NCSs with multiple sensors,actuators and controllers are connected with a network is presented in this dissertation.Because of the limitation on bandwidth, only a portion of the sensors and actuators can receive the information delivered by the controller.However,the other sensors and actuators have to wait.Therefore, the performance of NCSs not only depends on the controller designed,but also depends on the media transmitted efficiently.In terms of the feasible solutions to LMI,a necessary and sufficient condition is derived via the usage of networks in the context of decentralized control,and the set of periodically time-varying（PTV） local controller can be derived.（3） The problem of controller design for NCSs via digital communication is addressed. The systems under consideration are stabilized via state feedback,where the effects of sampled signal,state quantization,network-induced delay and packet dropout are considered.The proposed delay-dependent stability criteria are formulated in the form of LMIs method,which ensures asymptotic stability and a prescribed H_∞performance level for NCSs with admissible uncertainties.Maximum allowable delay bound of NCSs is obtained by solving a convex optimization problem.At the same time,quantized feedback stability criterion of NCSs with nonlinear perturbation is proposed such that the closed-loop NCSs are stable.（4） The problem of the quantized dynamic output feedback controller design for NCSs is proposed.By using the quantized information of the system measurement output and the control input,a novel NCS model is described.This model covers many network-induced features,such as multi-rate sampled-data,quantized signal,time-varying delay and packet dropout.Based on Lyapunov-Krasovskii functions, a less conservative stabilization criterion for the existence of a dynamic output feedback controller is derived by introducing slack matrix variables.The quantized control strategy involves the updating values of the quantizer parametersμ_i（i=1,2） （μ_i（i=1,2） take on countable sets of values which dependent on the information of the system measurement outputs and the control inputs）.Design of controller only needs the values of quantizers q_μ1（y_p（t））,q_μ2（y_c（t）） in the sample sequence t=s_i and t=σ_i（i=1,...,∞）.In terms of the feasible solutions to LMI,maximum allowable delay bounds of time-varying delay are obtained.Compared with the existing results,the merits of the obtained results are less conservative.（5） The problem of the robust H_∞filtering for uncertain NCSs,where measurement quantization,signal transmission delays and data packet dropout are considered, is addressed.The signal transmission delays are assumed to have both an upper bound and a lower bound,which is more practical than existing results for NCSs.Based on Lyapunov-Krasovskii functions,a less conservative stability criterion is proposed such that the closed-loop NCS is asymptotically stable with an H_∞performance bound.The quantized control strategy involves the updating values of the quantizer parametersμ（μtake on countable sets of values which dependent on the information of the system measurement outputs）.Design of filter only needs the value of quantizerq_μ（y_p（t）） in the sample sequence t=s_i（i=1,...,∞）.The H_∞optimal filter is formulated as a convex optimization problem.Moreover,the resulting criterion is further extended to more general cases,where the parameter uncertainty of system matrices is assumed to be of the polytopie type.更多还原