【作者】 马长林；

【导师】 方华京；

【作者基本信息】 华中科技大学 ， 控制理论与控制工程， 2006， 博士

【摘要】 随着网络技术的快速发展,越来越多的控制系统通过通讯网络实现闭环控制,控制系统的各部件之间的信息交换和控制信号的传递通过网络来完成,这就是本文的研究对象——网络化控制系统。在网络化控制系统的实际工程应用中,由于通讯网络的存在必然导致一些不确定因素的产生,其中起主要干扰作用的就是网络诱发的时延,在网络化控制系统中,时延具有随机性和复杂性,通常情况下它都数倍于采样周期,具有多步性,它的存在将降低控制系统的性能,甚至会使得系统变的不稳定,这使得网络化控制系统在工程应用中难以控制,严重时甚至会导致整个控制系统的崩溃而造成巨大的经济损失。因此,研究具有多步时延的网络化随机控制系统的控制分析和综合问题具有重大的理论和应用价值。本文对多步时延的网络化随机控制系统的控制分析和综合问题进行了深入的研究,提出了一些新方法,较好地解决了多步时延的网络化随机控制系统的建模、时延的统计特性研究、最优状态估计、综合控制分析和稳定性等问题。本文系统全面地介绍了当前国内外网络化控制系统的研究现状,分析了其主要研究方法,阐述了研究网络化控制系统的重要理论和实际意义;概括了国内外随机控制问题的研究成果,总结了其研究方法和基本思想;在结合随机控制和网络化控制系统二者研究的理论成果的基础上,提出了本文对多步时延的网络化随机控制系统进行研究的行之有效的方法。文中提出了两种新的控制模式,在新的控制模式下,一个采样周期内到达执行器的多个控制量更为有效的作用于受控对象,提高了系统的性能。在传感器、控制器和执行器的不同驱动模式下,考虑到随机延迟的影响,建立了四种不同的随机数学模型。本文针对不同的网络化随机控制系统模型,讨论了时延的统计特性和界限,利用Markov链理论对时延的随机性进行讨论,对两种新的控制模式下随机数学模型给出了关于Markov状态转移矩阵的新定理,并在此基础上给出了使系统保持稳定性的最大允许延迟界限的求解新方法;考虑到时延的随机性和复杂性,对于特别复杂的无法描述其概率分布和界限的网络化随机控制系统的情况,给出了一种时延在线估计的方法。本文采用几种典型的控制方法对网络化随机控制系统进行控制分析,主要有LQG最优控制法、无穷确界最优控制法、有限变确界最优控制法、基于模型预测的补偿控制法和时滞补偿控制法,特别是对存在空采样以及数据拒绝的网络化随机控制系统,使用基于模型预测的补偿控制法对系统进行补偿;另外还研究了各种控制方法下系统的稳定性问题,分析了各种控制方法对系统性能的影响,对网络化随机控制系统的稳定性进行了深入分析,构建了使系统稳定的控制器。对于无法获得完全状态信息的网络化随机控制系统,本文讨论了系统的最优状态估计问题,设计了系统的最优状态估计器和控制器,并证明了系统满足分离定理。在综合各种控制方法的特点的基础上,本文提出了一种网络化随机控制系统的综合控制法——基于α-Credibility的转换控制方法,其基本思想是通过对系统中多步时延的概率分布研究,根据时延落入的不同置信区间来采用不同的控制方法,这种分时区控制方法对系统性能有显著提高,达到了更好的控制效果;基于α-Credibility的转换控制法对网络化控制系统的智能化程度有一定的要求,解决方法是采用在基于Intelligent Agent技术的计算机网络化控制系统平台上进行转换控制,本文提出了一种基于IA技术的计算机网络化控制系统,并讨论了系统的可靠性问题。更多还原

【Abstract】 With the rapid development of network technology, more and more control systems adopt communication network to realize close-loop control, which means information exchange and control signal transmission among components are completed by communication network. This is the research object in this dissertation—networked control system. In engineering applications and practice, the insertion of communication network in the feedback control loop causes some indeterminate factors among which network-induce time delay mainly affects the system. In networked control system, time delay is stochastic and complex. In general, it is longer than one sampling period. Multi-step time delay can degrade the performance of control system and even destabilize the system, which gets the control of networked control system difficult in engineering application and will probably bring about enormous economic loss. Therefore, it is of great theoretical and applicable significance to research control analysis and synthetic problems. In this dissertation, these problems are deeply studied, and some new methods are presented to perfectly solve these problems, such as establishing models, the stochastic characteristics of multi-step time delay, optimal state estimate, synthetic control analysis, stability analysis, etc.In this dissertation, the present research situation of networked control system is extensively introduced. The internal and international research results, thoughts and methods of stochastic control and analysis are summarized. Then both theoretical results are combined, and suitable research approaches for stochastic networked control system with multi-step time delay are obtained. Two kinds of new control mode are presented, under which the system can make full use of control information. With the different driven modes of sensor, controller and actuator, four different mathematical models are established.According to different models, based on Markov chain theory, the stochastic characteristics and upper bound of time delay are discussed. New results of Markov transition matrix of two new control modes are presented. A new method to obtain maximum allowable delay bound is developed for stochastic networked control system with multi-step delay. An online evaluation method for networked control system with unknown probability distribution and bound is also introduced.Several typical approaches are adopted to realize the control analysis of stochastic networked control system with multi-step delay, including LQG optimal control method, infinite horizon optimal control method, finite moving horizon optimal control method, model predictive control method and time-varying delay compensation method. The model predictive control method is especially introduced to deal with vacant sampling and data rejection in networked control system. The stability analysis of control approaches above is also discussed. For networked control system with partial state information, the optimal state estimate is investigated. The optimal state estimator and controller are designed, which is presented in separation theorem.On the basis of various control analysis approaches above, a novel switching control method based onα-Credibility is provided. The controller of networked control system switches different control law to control the system according to time delay in different credibility intervals. The switching approach can improve the performance of networked control system. Computer networked control system based on intelligent agent technology to realize switching control method based onα-Credibility is presented, and the reliability of the system is discussed.更多还原