Analysis And Control Of Networked Control Systems

A great deal of attention has recently been focused on a class of networked control systems(NCSs) wherein the control loops are closed through communication networks. This family of systems is an integration of plants, sensors, controllers, actuators and communication networks of certain local field. However, the introduction of communication networks in the control loop makes the analysis and design of NCSs complex due to random network-induced delay, jitter, packet dropout as well as limited bandwidth. The modeling, analysis and control of NCSs are studied in this dissertation. The main contents are concluded as follows:(1) The modeling of the NCSs with time-delay shorter than one sample period is studied. The shortcoming of modeling method in current literatures is analysed and a novel method is proposed to model the NCS with time-driven sensor, event-driven controller and actuator as a class of discrete linear system with parameter uncertainties. Based on the Lyapunov theory and Matrix inequalities, the sufficient condition of asymptotic stability for NCSs is derived under the state feedback and output feedback condition, and the sufficient condition of the existence of guaranteed cost control law is also derived. The design methods of state feedback controller, output feedback controller, state feedback guaranteed cost controller and output feedback guaranteed cost controller are presented. Robust model predictive control method is used to design state feedback model predictive controller for NCSs.(2) The modeling and control problem of Multi-packet transfer NCSs(MNCSs) with uncertain time delay is studied. Supposing the sensor node uses multi-packet transfer, a plant model is used in controller node to evaluate the unreceived plant state, and the corresponding model state is updated by received plant state, then the state equation of NCS is derived. When the sensor nodes use static scheduling strategy, the MNCSs are modeled as uncertain periodic system. The MNCSs are modeled as discrete switched system when the sensor node using dynamical scheduling strategy. Based on the periodic system theory and matrix inequalities, the sufficient condition of asymptotic stability for uncertain periodic NCSs subjected to H_∞performance is derived, and the design methods of H_∞controller and guaranteed cost control are given. Output feedback MNCS is studied, a state observer is introduced in controller node, and periodic system theory and switch system theory with respect to static scheduling and dynamical scheduling strategy is used to get the sufficient condition of asymptotic stability for output feedback MNCS and the design approaches of state observer and controller.(3) Analysis and design problem of NCSs with data packet dropout and long time delay is studied. Consider NCSs with output delay and data packet dropout, in order to compensate the influence of time delay and data packet dropout, a state predictor is designed in the controller node. Design methods of D-stable state predictor and controller are presented. Consider NCSs with control delay and data packet dropout, a buffer is introduced to convert time-varying delay to fixed delay, the sufficient condition of asymptotic stability and the controller design method for such NCSs is putforward. For NCSs with data packet dropout, output delay and control delay, combined with state predictor, the NCS is modeled as a class of asynchronous dynamical system constrained with event rates. The sufficient condition of exponential stability for NCS with known event rates is derived, and the controller design method is given.(4) Adaptive predictive control problem of NCS with data packet dropout and network induced delay is studied. An appropriate queuing scheme is used to compensate the effect of data packet dropout. Let actuator has the ability of sending data packet, and every packer has time stamp, the capacity of data packet is utilized sufficiently, at the sample time, actuator and sensor send current and certain past samples of control input and output series, when controller node receives data packet, according to the time stamp, the data vectors is construct to identify the model parameters. Identification result is used to predict process output series up to current time and then compute control series. The application of the proposed method in nueral netork based adaptive predictive control for nonlinear NCS is also studied.