【作者】 邓玮璍；

【导师】 费敏锐； 李慷； George W.Irwin；

【作者基本信息】 上海大学 ， 控制理论与控制工程， 2012， 博士

【摘要】 网络技术的发展同时带动了许多新学科方向的兴起,其中网络控制系统就是兴起不久并受到极大关注的学科方向之一,它是涵盖控制、通信、网络和计算机等多种学科的交叉方向,并且有着广泛的实践应用。随着广大科研人员对网络控制系统不断扩展和深入的研究,在理论和应用上都已经取得了许多满意的成果。但是仍然存在许多开放的问题有待解决。从网络控制系统的稳定性与控制器设计受到各种条件制约或限制的角度出发,本文主要针对通信受限、被控对象信息完备性受限及控制系统组件正常运行受限这三个条件,研究了在相应条件下及各种混合条件下网络控制系统稳定性和控制器设计理论。主要研究成果如下:首先,考虑理想情况,即被控对象信息完备,对象模型已知,并且只考虑单通道通信时间受限情况下的网络控制系统稳定性分析与控制器设计。这一部分主要研究了两种情形:一是针对线性对象模型及Markov随机时滞模型,使用Lyapunov-Razumikhin方法对网络控制系统渐近稳定性进行分析,设计了动态输出反馈控制器,保证了系统的镇定。二是考虑复杂非线性对象模型、给定上界的时变时滞模型,使用Lyapunov-Krasovskii方法对网络控制系统渐近稳定性进行了分析,设计模糊控制器,保证了系统的镇定。其次,考虑较为实际的情况,即在被控对象信息不完备,具有一个不确定范围且不确定界是已知的情况下,并考虑单通道通信时间受限的网络控制系统稳定性分析与控制器设计。在第一个问题的两种情形基础上,分别加入对象不确定成分,并采用范数有界不确定模型形式刻画对象不确定部分,得出了网络控制系统渐近稳定条件,并设计了相应的控制器。针对线性情形下由不确定性因素诱导的双线性问题,采用了一个经修改后的锥形补偿线性化算法对其进行了线性化求解。针对非线性情形,使用模糊系统的思想并基于平行分布补偿技术对网络时滞进行了建模,这种时滞模型更符合实际应用研究。第三,考虑更为实际的情况,即被控对象模型完全未知的情况下,并考虑一类单通道通信时间受限的无线网络控制系统稳定性分析与镇定及跟踪控制器设计。在这种情形下,采用首先进行对象模型的辨识,然后将辨识好的模型用于控制的策略。针对一般的Hammerstein系统,采用一类逆高斯网络时滞统计模型刻画无线网络延时,提出了一个基于模型的网络迭代辨识方法,利用辨识好的模型得到Hammerstein系统非线性部分的逆函数去处理非线性,使用H∞方法抑制由模型误差带来的扰动,并使用基于模型的控制器对延时或丢失的数据包进行重新估计。通过以上一系列方法,可以保证Hammerstein系统在无线网络控制下达到均方渐近稳定。第四,考虑通信时间受限的多通道网络控制系统稳定性分析与控制器设计。前面三个问题主要是考虑了被控对象信息完备性受限程度递增,单通道通信时间受限的网络控制系统的稳定性和控制器设计问题。从这个问题开始到下一个问题,本文将转向通信受限程度递增,对象模型假设已知的网络控制系统稳定性分析和控制器设计。本问题假设多个传感器通过多个通道接入令牌环网络,分别使用一个Markov链和多个Markov链刻画其随机时滞特性,建立参数依赖的Lyapunov函数分析系统稳定性,设计状态反馈控制器并结合缓存技术对丢包进行补偿,保证了基于多通信通道时间受限的网络控制系统稳定和镇定。第五,考虑通信时间与通道信噪比均受限情况下的网络控制系统稳定性分析与控制器设计。首先同时考虑单通道信噪比与时变时滞,使用Lyapunov稳定理论和线性矩阵不等式方法得到连续时间网络控制系统镇定所需要的最小信噪比。然后同时考虑单通道信噪比与Markov链形式的随机时滞,使用同样的方法得到离散时间网络控制系统镇定所需要的最小信噪比。最后考虑一类异构网络情形,即一侧是受通道噪声影响较大的单通道无线网络,另一侧是受通信时滞影响较大的多通道令牌环网络,仍然使用Lyapunov稳定理论和线性矩阵不等式方法可以得到此类网络镇定所需要的最小信噪比,并且给出信噪比、对象不稳定极点与网络时滞的一个关系式。第六,考虑执行器故障的网络控制系统稳定性分析与控制器设计。此问题是从控制系统组件正常运行受限的角度展开研究的,而前面五个问题都是在假设执行器、传感器无故障条件下对网络控制系统稳定性及控制器设计进行的研究。从可靠性设计的实际考虑,控制系统组件出现故障是必须考虑的问题。所以这个问题考虑了执行器有部分出故障,通信时滞时变且有上界的连续时间网络控制系统稳定性,并且设计了自适应故障容错控制器,以保证系统在故障和通信时间受限情形下可以正常运行。在上述六个问题的理论分析和设计完成之后,均进行了数值仿真验证,表明了本文方法的正确性和有效性。另外在做无线网络时滞和通信信噪比的数值仿真时,结合了英国女皇大学智能系统与控制课题组通过实验得到的逆高斯分布的时滞统计模型和上海大学电站自动化技术重点实验室的双并联倒立摆装置模型,为所研究问题能进一步应用到实际提供了依据与基础。更多还原

【Abstract】 The rapid development of network technology has advanced many new research areas, among which the networked control system (NCS) has drawn a lot of attentions in recent years. The NCS which covers multiple disciplines, including control, communication, networks and computer, has been widely applied in many engineering fields and a number of theoretic achievements have also been reported. Yet, some open problems still exist to be solved. Among various issues restricting the application of NCSs, the stability and controller design have been studied in this thesis, considering various communication constraints, completeness of plant information, and operation conditions of control system components. The following technical contributions have been made in this thesis:1) First, we assume that the information of plant is complete, i.e. the model of the plant is known, a single communication channel presents in the system and the communication time is limited. Given these assumptions, the stability analysis and controller design are examined under two cases: in the first case, Lyapunov-Razumikhin approach is used to analyse the asymptotic stability of NCSs with a linear plant model and Markov type stochastic delays, and a dynamic output feedback controller is designed to stabilize the system. In the second case, the Lyapunov-Krasovskii method is employed to analyse the asymptotic stability of NCSs with a more complex nonlinear plant model and time-varying delays with an upper bound, and a fuzzy controller is designed to deal with the system stabilization problem.2) Then, we consider the problem where the information of the plant is incomplete, and assume that the plant model has an uncertainty with known boundary. Following the two cases studies in 1) we can obtain the stability conditions and the corresponding controller design with norm bounded uncertainties. Further, a modified cone-complementarity linearization (CCL) is used to solve bilinear problems and a parallel distributed compensation (PDC) technique is employed to model the network delays.3) Next, we assume that the plant model is unknown a priori, which is more common in many engineering applications. In particular, we investigate the identification and control of Hammerstein systems over wireless networks. Firstly, a new model-based networked identification (MBNI) method is proposed for Hammerstein systems. Then with this identified model, model-based control strategy with inverse transformation is applied to networked control to compensate network induced delays and nonlinearity respectively. In addition, H∞control is used to deal with disturbances induced by the error between the model and the real system. Then a sufficient condition for system mean-square asymptotic stability is obtained in the form of linear matrix inequalities (LMIs) and controller gains can thus be directly solved.4) We then consider the NCS with multiple communication channels and stochastic delays. In 1)– 3), we focus on stability analysis and controller design with different degree of completeness of plant information and a single communication channel. In this part, we concentrate on different communication constraints. We assume that multi-sensors are connected in a Token-ring network, and one and multiple Markov chains are used to characterize the stochastic delays. A parameter-dependent Lyapunov function is employed to analyse the system stability, and the state feedback controller and the buffer technique are applied to the system stabilization problem.5) We further investigate the stability and stabilization of NCSs subject to both communication time and signal-to-noise ratios (SNRs) constraints. Firstly the minimal signal-to-noise ratios (SNRs) required for stabilisability are obtained in continuous-time networked control system (NCS) with varying delays using the linear matrix inequality (LMI) approach. The results are then extended to discrete-time heterogeneous networked control systems consisting of a wireless network with more channel noise and a Token-ring network with more delays. Finally the above theoretic results are tested on several numerical examples, and relations among unstable poles, delays and SNRs are illustrated explicitly.6) Next, we consider the situations where the operation conditions of control system components are subject to constraints. In 1)-5) it is assumed that the actuators and sensors are fault free, but these components sometimes may be subject to faults. To tackle this problem, an adaptive fault-tolerant control approach for networked control systems (NCSs) with actuator faults is developed. Based on a new network-induced delay model proposed recently, we design an adaptive state feedback controller. A sufficient condition for the existence of a controller is given in terms of linear matrix inequality (LMI), which guarantees the stability of NCSs under normal and faulty conditions in the H∞sense.7) Finally, the validity and effectiveness of the above techniques are demonstrated in numerical simulations. It should be noted that in the simulations of wireless network delays and SNRs, the inverse Gaussian distribution delays model developed by the Intelligent System and Control group at Queen’s University Belfast, and the two inverted pendulums on carts coupled by a spring used in Shanghai Key Laboratory of Power Station Automation Technology at Shanghai University, have provided a solid foundation for development and application of the proposed techniques.更多还原