Performance Analysis And Robust Control For Stochastic Networked Control Systems

Networked control systems (NCSs) are special distributed control systems in which feedback control loops are closed via communication networks. Comparing with the conventional point-to-point control systems, NCSs present a number of advantages, such as low cost,reduced system wiring, ease of system diagnosis and maintenance, and increased system agility. The targeted applications are industrial plants, transportation, power gird, and unmanned aerial vehicles (UAVs), etc. However, the introduction of communication networks in feedback control loops complicates the analysis and synthesis of NCSs, because the network-induced transmission delays, data packet dropouts and fluctuation or limitation of bandwidth will inevitably degrade the control performance of NCSs, or even cause the system to be unstable. Consequently, the research on NCSs is of great theoretical and applicable significance and has appeared to be a topic of significant interest to the control community. In this dissertation, the modeling, analysis, and control of NCSs with uncertain network quality, such as network-induced delay, packed dropout, and limitation of bandwidth are deeply studied by applying the robust control and stochastic control theory. In terms of the Lyapunov approach and linear matrix inequality (LMI) techniques, stability conditions and design methods of the robust controller are presented for NCSs with uncertain network quality. By introducing kinds of optimal performance index, the inner relationship between performance index and network quality is also established. The main contents are as follows:The observer-based guaranteed cost control problem is first studied for a class of NCSs with random data packet dropouts. Two kinds of packet dropouts (the packet dropout from sensor to controller (S/C) and controller to actuator (C/A)) are both modeled by two mutually independent stochastic variables satisfying the Bernoulli binary distribution. New NCS models are provided considering both single- and multiple-packet transmission. A new observer is designed to estimate the states and system input simultaneously by constructing an augmented system in which the original system input is regarded as a new state. Based on this observer, a guaranteed cost controller is designed such that the closed-loop system is stochastically exponentially mean-square stable and the cost function value is not more than a specified upper bound. Then the design method of such controller is formulated in terms of the LMI approach and iterative Cone Complementarity Linearization (CCL) algorithm.The observer-based H∞control problem is investigated for a class of discrete NCSs with communication constraints. Assuming that the network has limited capacity and the total bandwidth is random fluctuated satisfying a known probability distribution, the NCS models are provided considering both single- and multiple-packet transmission. The sufficient conditions on the stochastic stability of NCSs are obtained in terms of different bandwidth distribution and communication constraints. By introducing the optimal H∞performance index, the inner relationship between H∞performance and communication constraints is established. Using the iterative LMI approach and CCL algorithm, the H∞controller is designed such that the closed-loop system is stochastically mean-square stable and the optimal H∞disturbance attenuation level is also achieved.The state-feedback H∞control problem is investigated for a class of NCSs in the discrete-time domain with random but bounded delay. At each sampling time, the current S/C delay is known for controller by, e.g., the time-stamping technique. Considering both single- and multiple-packet transmission, the closed-loop NCSs are modeled as stochastic discrete-time jump linear systems with the S/C time delays are modeled as a Markov chain. In order to reduce the conservativeness of the results, the state-feedback controller gains are switched depending on the value of the S/C time delays. The sufficient conditions on the existence of H∞controller are obtained. The iterative LMI approach and CCL algorithm are employed to calculate the state-feedback gains.Finally, the Sampled-data H∞control problem for a class of networked control systems (NCSs) with random packet dropouts is investigated. The NCS is modeled as a sampled-data system which involves a continuous plant, digital controller, event-driven holder, and network channels. In this model, two types of packet dropouts in the S/C and C/A sides are both considered, which are described by two mutually independent stochastic variables satisfying Bernoulli binary distribution. By applying an input delay approach, the sampled-data NCS is transformed into a continuous time-delay system with stochastic parameters. An observer-based control scheme is designed such that the closed-loop NCS is stochastically exponentially mean-square stable and the prescribed H∞disturbance attenuation level is also achieved. The controller design problem is transformed to a feasibility problem for a set of LMIs.