Analysis And Synthesis For A Class Of Networked Control Systems With Variable Quantization Density

With the rapid development of network technology,networked control systems have become one of the hottest research topics in both academic and industrial com-munities.Networked control systems integrate automatic control,computer technology and communication technology with feedback loops being closed through the network,which have wide application prospects.On one hand,the introduction of network helps networked control systems break through the limitations of property and function of traditional control systems.On the other hand,network also brings difficulties to stability analysis,system performance analysis,design and synthesis of networked control systems.In practice,signal quantization occurs in both control input and measurement output channels,and the capacity of network may vary with the current state of the network.This dissertation focuses on the in-depth studies of signal quantization in a class of networked control systems in the scenario of variable communication capability,which can make a balance between the system performance and the amount of transmitted data.Chapter 2 addresses the problems of stability analysis,H_? performance analysis and H_? controller design for a class of networked control systems with disturbance input,and signal quantization occurring in both control input and measurement output channels.Two sets of different logarithmic quantizers with variable quantization densities are utilized to quantize signals in the above-mentioned two channels.It is considered that the variable quantization densities of two logarithmic quantizers are selected separately from two sets of quantization densities,which are governed by a Markov chain.With the aid of system-mode-dependent and quantization-error-dependent Lyapunov function as well as the method of convex combination,we arrive at a set of mode-dependent H_? controllers that guarantee the stochastic stability and the prescribed H_? performance of the closed-loop system with less conservation.Chapter 3 investigates the output-feedback control problem of a class of networked control systems under the circumstance of the system state cannot be fully measured.It is considered that the quantization densities of the logarithmic quantizers in both control input and measurement output channels are variable by taking values in two finite sets,which can be modeled by a semi-Markov process.With the usage of semi-Markov kernel approach to modeling the closed-loop semi-Markov system with variable quantization densities,and based on the system-mode-dependent and quantization-error-dependent Lyapunov function,sufficient conditions ensuring?-error mean square stability of the closed-loop system can be obtained,by assuming the upper bound of each mode's sojourn-time exists.Besides,a class of mode-dependent output feedback controllers are designed to guarantee the?-error mean square stability of the closed-loop system by utilizing the cone complementary linearization algorithm to solve the nonconvex problem.Chapter 4 presents the issue of the H_? control for a class of networked control systems with fixed quantization densities and continuously variable quantization densities separately.A more general problem for multi-input and multi-output control systems is that all the components of the control input and measurement output are quantized via different logarithmic subquantizers.In section 4.1,the quantization densities are considered to be fixed,and with the approach of convex combination,sufficient conditions for stability and H_? performance of the closed-loop system are derived,then a model-based state feedback H_? controller can be achieved.Considering the variable quantization densities may vary continuously,section 4.2 proposed a model-free method to obtained the algorithm of optimal H_? controller with the approach of adaptive dynamic programming.The optimization problem is to find a control strategy that minimizes the l₂ gain from disturbance input to measurement output,and under arbitrary disturbance input,the closed-loop system is stable with a guaranteed H_? performance index.By defining a quadratic function to approximate the H_? performance function,we can calculate the quadratic matrix and H_? controller gain by using the system input and response data without knowing system matrices,then a model-free control strategy can be derived.