Design Of Scheduling Protocols And Treatment Of Time-delay In Networked Control Systems

In order to handle the spatially distributed large-scale plants and remote plants, people could use the existing communication network to enable all the components, including sensors, actuators and controllers, to exchange information with each other. The introduction of a network cancels the complicated routing, leading to simpler configuration, easier maintenance and higher reliability, etc., together with some other adverse e?ects.This dissertation is devoted to the treatment of the communication constraint, under which not all the information of each component can be transmitted, and the time-delay.On one hand, we design scheduling protocols, which designate the transmission instants and pick out the nodes to grant the right to access the network, in hope of satisfying the communication constraint condition. On the other hand, we use the model-based architecture to reduce the amount of the information transmitted via the network. The time-delay, composed by the time the nodes spend to await its turn to transmit information and the time cost by transmission, is ubiquitous in network. For the communication constraint, we design scheduling protocols. For the time-delay, we discuss a method to improve Lyapunov’s Second Method. The main contents are as follows.First, we design Contain-the-Largest-State-to-(CLS-) scheduling protocols for a group of systems running under communication constraint. The network, which is modeled as a group of on-o? switches, links the controllers and the plants. Under the communication constraint, only one switch can be closed, that is to say, only one pair of controller and plant can be linked at any time. To select the proper system, we discuss the CLSprotocol, which picks out the plant with the largest state and links it to its controller. The norm is used to depict the size of the state. The existing result is based on the weighted norm of the state. We consider the CLS- protocol based on any norm. Though all norms are equivalent in discussing the asymptotical stability of the system, di?erent norms lead to di?erent scheduling protocols.Second, we design Maximum-Error-First(MEF) scheduling protocols in time-triggering style. The signal is transmitted at discrete time instants, which are to be designated, and is held constant by a zero order holder until it is updated. The error of each node is defined as the di?erence between the current state value and the value transmitted via the network last time. The node with maximum error is picked out and permitted to transmit its information via the network, while other nodes must wait for the next competition.The transmission instants are determined through a time-triggering style. The shorter the intervals between transmission instants are, the less the communication resource is consumed. Based on the existing results, we give a better estimate of MATI, and demonstrate the advantages of the given result through some simulation.Thirdly, we design scheduling protocols in event-triggering style. Besides this, in order to reduce the network communication, a model used to estimate the state of the plant is contained in the controller side. The estimate given by the model is used to calculate the control signal substituting the actual value of the plant state. Only at discrete instants,the model state is updated using the plant state transmitted via the network. The update instants are determined in event-triggering style, with the event chosen to be that the error grows to a prescribed level. We analyze the simultaneous stability of a group of systems mounted on the network. We investigate the case where the plant state, used to update the model state, could not be wholly transmitted due to communication constraints. We give the method to design the triggering event, and analyze the stability of the system.Fourthly, we treat the time-delay in networked control systems. We investigate the control of a classic system of two pendulums coupled by a spring, which is paradigm test-bed in distributed control. We analyze its stability and H∞ performance, and give a method to design the controller. The tool used is Lyapunov-Krasovskii functional. In calculating the time derivation of the functional, we have to deal with some integral terms.To guarantee the nonnegativity of the derivative of the functional, we have to estimate the integral terms. We use the method of matching the squares, and make comparison between this method and the method based on Jessen’s Inequality.