IV摘摘 要要 进入 21 世纪的控制系统将以网络为主要特征。自动化与控制需要更深层次地渗透通信与网络技术，现场总线等具有实时性的控制网络如今已广泛地进入工业控制领域，以智能网络节点为基础实现传感器、控制器、执行器的网络化控制系统结构将是新一代控制系统形式的大势所趋。网络控制系统由于网络的介入，不可避免地带来了许多问题：如网络传输导致的反馈和控制输入中的时滞、网络带宽的限制使得数据必须分批传送、通信过程失败造成的分组丢失、不同的网络节点之间系统时钟的异步等等。这些问题的存在，不但会降低系统的控制性能，而且还是引起系统不稳定的潜在因素。本文在详细分析这些网络化的不利影响的基础上，利用李亚普诺夫稳定性理论、时滞系统理论、切换系统理论、脉冲系统理论的一些现有结果，对网络控制系统的建模、稳定性和控制方法进行了相关研究。 为确保建模的合理性与实用性，本文首先概括总结了当前网络控制系统的研究文献中关于网络化的影响分析，以及解决这些不利影响的各种现有方法。定性的网络化影响分析对于网络控制系统的建模工作具有重要的指导意义。 网络传输时滞是网络化给控制系统带来的最显著的不利影响，迄今已有的工作多数只针对定常网络传输时滞的研究，关于时变时滞的讨论则非常少见。为填补这个空白，本文对时变传输时滞影响的网络控制系统进行了建模与分析工作。首先在离散系统的框架下为此类系统建立了时变时滞线性区间系统模型，进一步利用离散系统的李亚普诺夫稳定性理论与方法以及矩阵不等式的相关理论，研究了这些模型的鲁棒稳定性问题；另外，考虑到非线性被控对象在实际系统中具有更广泛的代表性，本文还研究了非线性不确定时变时滞网络控制系统的建模及稳定性分析工作，利用非线性映射的雅各比矩阵等手段得到了简洁的稳定性判据。 除了时滞的影响之外，不可靠的网络通信时常会造成分组信息的丢失，亦即所谓的数据丢包现象。对于数据丢包为主要不利影响的所谓有损网络控制系统，本文分别从连续和离散两种情况使用切换系统模型进行了描述，并首次提出了这类切换系统模型与一类脉冲系统的关于稳定性的等价定理，从而将网络控制系统转换为脉冲系统进行研究。我们将切换系统中关于“慢切换”情况下驻留时间与稳定性关系的结论予以引申应用于脉冲系统之上，得到了此类脉冲系统(含脉冲微V分系统和脉冲差分系统)的渐近稳定性条件。注意到这些条件蕴涵着丰富的物理意义，与实际情况基本符合。另外，对于离散情形的有损网络控制系统，本文利用研究得到的脉冲差分方程稳定性条件和线性矩阵不等式的性质，进一步地讨论了其在参数为不确定性时的鲁棒镇定问题。这些模型、稳定性条件与提出的控制律设计方法都是首创。 关于数据丢包影响的有损网络控制系统，本文还研究了它们在模型为不确定时满足某二次型性能指标的保成本控制问题。由于连续线性时变系统的基本解矩阵显式表达的困难性，本文并未直接研究在连续情况下的有损网络控制系统的保成本控制问题，而只是分析了其中的一个特例一类参数不确定脉冲微分系统的保成本控制问题。基于线性矩阵不等式优化方法，得到了其保成本控制律的存在条件和设计方法。关于离散情形，直接利用已经得到的鲁棒镇定问题的结论以及线性矩阵不等式的方法，我们同样得到了有损网络控制系统的保成本控制律的存在条件和设计方法。 另外，对于不确定有损网络控制系统，我们还研究了此类系统在某种程度的数据丢包率影响下具有一定 H性能界的鲁棒状态反馈控制问题，得到了其控制律的设计方法，结果以一组线性矩阵不等式表示。 本文首次提出了网络化脉冲控制系统的概念并研究了新环境下的系统稳定性问题。针对数据丢包的情形，本文定义了一个 N 步丢包率的概念，并论证了 N 步丢包率与网络化脉冲控制系统渐近稳定性之间的关系。考虑传输时滞对网络化脉冲控制系统的影响，我们构造了基于系统模型的预测器，并使用此预测器根据延迟的反馈值预测出系统当前状态的估计值，然后由此估计值计算得到控制输出。对于这种控制系统形式，我们还得到了其渐近稳定的一个充要条件。 关键词关键词：网络控制系统 切换系统 脉冲系统 时滞 不确定性 稳定性 鲁棒镇定 保成本控制 VIAbstract One of the main features of modern control systems is the data transmission via networks. Automatic control technology is involved with communication network more and more. Various real-time networks such as field-bus are extensively applied into industrial control engineering. The networked control system structure will prevail in the future, whose sensors, controllers and actuators are all connected via networks. With the impact of network circumstance, many issues emerged inevitably, such as network induced transmission delay, multi-channel transmission brought by the limited bandwidth, data packet dropouts from the failed communication, the asynchronous clock among network nodes etc. These problems not only depress the performance of normally designed control systems, but also destroy the system stability. After analyzing the disadvantages brought by network, this paper investigated the modeling, stability anaysis and control law synthesiss of networked control systems. The related theories and methods include Lyapunov stability theory, delay system theory, switched system theory and impulsive system theory. First of all, this paper analyzed detailedly the effects brought by the network and presented many methods to cope with the harmful factors. Though the analysis is basically qualitative, it will help model networked control systems. The transmission delay is the most apparent disadvantage from the network connectivity. A large amount of materials studied the constant transmission delay, yet seldom considered the time varying ones. To supply this gap, we modeled the networked control systems with time varying transmission delays and obtained some results. First we established the linear discrete interval system model with time varying delays for networked control systems. Then the robust stability of this model was studied via discrete Lyapunov stability theory and matrix inequality methods. Otherwise this paper established nonlinear networked control systems model with time varying delays and investigated its exponential stability problem. The stability condition is obtained in simple matrix inequalities by applying the Jacobin matrices. The data packet dropouts event is also one of the most important phenomena arising from network connectivity. This paper described the networked control systems with data packet dropouts by applying the switched system model in both continuous and discrete cases. Equivalent impulsive system models (including both impulsive differential system and impulsive difference system) were then presented for the first time. So we can study the networked control systems based on these impulsive systems. We extended the stability results of switched systems with respect to dwell-time to these VIIimpulsive systems. The asymptotic stability condition of equivalent impulsive system models was then obtained. Note that the physical meaning of the stability condition is obvious. Our conclusion fits the reality very well. On the other hand, the robust stabilization problem was discussed by applying linear matrix inequalities and the stability condition of impulsive difference systems was obtained. The models and methods are all created for the first time. Furthermore, the guaranteed