Analysis And Design Of Networked Control Systems With Random Long Delays

Networked control systems (NCSs) are feedback control systems wherein the control loops are closed through a shared communication network. The enormous advantages of NCS, including lower cost, reduced system wiring, increased system agility, ease of system installation and maintenance, etc., which has attracted more and more research interest by people. However, the insertion of the communication network in the feedback control loop brings many new issues inevitably, such as network-induced delay, data packet dropouts, communication constraints, etc., which makes the analysis and design of NCS exceptionally complex.Among so many issues of NCS, the network-induced delay is regarded as the major cause of the deterioration of system performance and potential system instability. It is because of the limited bandwidth, communication media with time division multiplex access and the different treatment of data packets, especially, in many cases, it cannot be ignored. Determined by the media access control protocols and topologies, the delays can be divided as constant delays, bounded delays, random delays, etc.. Compared with the constant delays and bounded delays, the random or time-varying delays are more universal in the actual control system and more difficult to dealt with. The characteristics of random and time-varying of the closed-loop NCS is derived by the random and time-varying delays. Thus, it is difficult to use conventional linear time-invariant system theory to analysis and design NCS. How to design an effective controller to overcome the effective of the random and time-varying delays has become an urgent problem to be solved.This thesis is concerned with the modeling, stability analysis and control problems of a class of discrete-time networked control system with random long delays by using robust control theory, Markov stochastic control theory and linear matrix inequality (LMI) techniques, etc.. The main contributions of the thesis are as follows:Firstly, the network-induced delays in the feedback channel and forward channel are modeled as two different chains. The relationships of the controller input delay and feedback channel delay, the plant control input delay and forward channel delay are given by timing analysis. Considering the state feedback control law, the resulting closed-loop NCS is modeled as a Markovian switched system which depends on the current and past time delay by the state augmentation techniques. By using the Lyapunov method and linear matrix inequality (LMI) technique, a sufficient condition for the closed-loop NCS to be stochastically stable is obtained. Then, the stabilizing controller design method is derived by using a cone complementarity approach. Moreover, an example is given to demonstrate the effectiveness of the proposed method.Secondly, a new modeling method is presented to avoid the calculation by expanding the state dimension. Based on the principle of using the most recent available data, the resulting closed-loop NCS is modeled as a Markovian stochastic time delay system which depends on the network-induced delays not only of current time instant but also past time instants. Sufficient conditions are derived for the closed-loop NCS to be stochastic stable and achieve a quadratic performance upper bound via Lyapunov method and LMI technique. A design procedure is also presented for the state feedback guaranteed cost controller. Finally, an illustrative example is given to demonstrate the effectiveness of the proposed method.Finally, based on the improved model, the closed-loop NCS is modeled as a Markovian stochastic time delay system with an external disturbance. The NCS with the state feedback controller has strong robustness by the H_∞control theory. With the stability and H_∞performance analysis, the H_∞state feedback controller is designed, which ensure the NCS to be stochastically stable with the anti-jamming capability. The effectiveness of the proposed method is illustrated by MATLAB simulation.