Filtering Of Network-based Switching Systems With Redundant Channels

Switched systems could be used to model the phenomenon of mode switching and the approach of switching control, which has been a hot research topic in the field of system and control. In accordance with the types of switching signals, it could be divided into stochastic switching and nondeterministic switching. The studies on them still exist a large number of problems which need to be found, proposed and solved, and some crucial research methods still need to be improved, although the researchers at home and abroad have obtained many important findings on the analysis and synthesis of such two typical switched systems. Furthermore, as the control system performance requirement which actual production needed is higher and higher, and the dynamic mechanism of control object is more and more complex, the switched system only considering simple subsystem dynamic usually could not well reflect many real dynamic processes, also the obtained results are difficult to apply to the practical system with complex dynamics,including packet dropouts, time-varying delays, quantization, etc. Therefore, the only way to solve the above issues is to further systematically and profoundly carry out the modeling, analysis and synthesis for stochastic and nondeterministic switched system in the networked environment, as well as the studies on such switched systems with complex subsystem dynamics, to analyze the quantitative or qualitative relationship between the complex dynamics and the system performance.This thesis aims to further profoundly deal with the filtering issues of network-based switched system including stochastic switching and nondeterministic switching, and to propose some new models, new problems and new approaches on the basis of the existing work. This thesis provides some new concepts, such as asynchronous filter, time-varying Lyapunov method, non-weighted performance index and so on, eliminates some ideal assumptions which are not accord with the practical engineering to reduce the conservatism of the developed results. Then, it deals with the generalized H2 asynchronous filtering and resilient H∞filtering issues of discrete-time Markov jump systems(MJSs). Then,to improve the reliability of data transmission, it proposes some new design methods of network-based system in the network environment, for instance, the network with single redundant channel, the network with multiple redundant channels and the network with variable communication capability, etc. The distributed filtering issue of discrete-time nonlinear MJSs, the generalized H2 filtering issues of discrete-time switched systems with modal persistent dwell time(MPDT) switching as well as the resilient H∞filtering issue of network-based system are investigated respectively. Besides, an example of cloud-based automotive suspension system is utilized to demonstrate the effectiveness of the developed approaches, also some necessary theory bases are provided for practical engineering.Chapters 2-3 first summarize the existing results and shortcomings on the topics. In terms of this, both the existing mode-dependent and mode-independent filters could not well reflect the case when the filter cannot detect the mode jumping of system completely, that is, as there exists delays in detecting mode jumping of system, the asynchronous phenomenon of mode jumping appears between the system and the filter. Therefore, the insufficient contained in the above two kinds of filters could be made up by the proposed asynchronous filter, which also enriches the results of filtering of MJSs. Firstly, in the networked environment without redundant channels, the generalized H2 asynchronous filtering issue is investigated for a class of discrete-time MJSs with random occurring nonlinearities(RONs) and time-varying delays. The RONs are used to model a class of sector-like nonlinearities that occur in probabilistic way according to a Bernoulli sequence. A monotonicity is disclosed in achieving a better filtering performance index while the degrees of asynchronous jumps are decreased. Then, the resilient H∞filtering problem is studied for a class of discrete-time MJSs considering time-varying delays, unideal measurements and multiplicative noises. Two independent Bernoulli sequences are used to describe the phenomena of random occurring quantization and packet dropouts.The quantitative relationship is analyzed between the system performance and the variation of additive gain disturbance, the multiplicative noises, and the rate of packet dropouts. A similar monotonicity is revealed in achieving a better filtering performance index as decreasing the degrees of asynchronous jumps. The idea of asynchronous filter proposed in this part provides a theoretical basis for the following distributed filtering issue of discrete-time MJSs with redundant channels, also points out the direction for further studying the analysis and synthesis of hierarchical hybrid systems.Chapter 4 is concerned with the distributed filtering issue of discrete-time MJSs,which is a very important filtering problem in the area of stochastic switched system. A single redundant channel is introduced to the measurement output model to improve the reliability of data transmission in the sensor network, and the corresponding distributed filtering issue is investigated for a class of discrete-time nonlinear MJSs. A stochastic switching topology structure is constructed to describe the variations of sensor nodes more precisely. The sufficient condition with less conservatism is derived using the Lyapunov function including the nonlinear information, and the effectiveness of introducing one redundant channel is demonstrated to obtain a better average H∞performance index. This part establishes a theoretical foundation for studying the asynchronous filtering problem of MJSs with various complex dynamics, also provides a theoretical reference for the following filtering issues of modal persistent dwell time(MPDT) discrete switched system with multiple redundant channels.Under the networked environment with multiple channels, Chapter 5 investigates the filtering issue of discrete-time nondeterministic switched system with MPDT switching.The PDT switching, which has the capability of describing a switched system with hybrid slow and fast switching feature, is more general than the dwell time switching and the average dwell time switching. The idea of multiple redundant channels is proposed to reduce the possibility of packet dropouts in the network, and the random measurement losses, occurring either in the primary channel or in some redundant channel, are modeled respectively by Bernoulli distribution and multinomial distribution. The non-weighted generalized H2 filtering issue is studied for a class of discrete-time switched system with MPDT switching and multiple redundant channels. The non-weighted form has stronger disturbance attenuation property compared with the weighted performance index. An example of discrete-time switched neural networks is illustrated to verify the advantage of multiple redundant channels and the mode dependence of PDT switching signal. This part provides a theoretical basis for the analysis and synthesis of nondeterministic switched system with PDT switching, also the proposed idea of multiple redundant channels provides a theoretical guidance for the following filtering problem of network-based system with variable communication capability.A typical application of using the switched system to model the switching control method is given in Chapter 6, and the resilient H∞filtering problem is investigated for a class of network-based system with variable communication capability. It is very possible that multiple channels coexist and the number of available channels is not fixed, leading to a variable communication capability that can be weighed in many ways including the packet arrival rate. The channel with variable numbers is introduced to the measurement output model, and the regularity of the variation is considered to be of the so-called MPDT property. By using the quasi-time-dependent Lyapunov function approach, sufficient conditions with less conservatism are obtained on the existence of resilient filter. Finally,an example of cloud-based automotive suspension system is utilized to demonstrate the effectiveness of the developed techniques, to analyze the quantitative relationship between the variation of additive gain disturbance and the system performance. This part represents the first attempt of applying the H∞filtering of network-based system with variable communication capability to the practical engineering problems, which extends the ideas of solving the filtering problem for network-based system with variable communication capability and provides a new theoretical guidance for practical engineering, also establishes a solid theoretical basis for further in-depth studying the analysis and synthesis issues of network-based system with variable communication capability.