Study On Robust Control Schemes Of Networked Control System And The Implementation Of Simulation Platform

The NCS is a feedback control system wherein the sensors, actuators and controllersas the network nodes connected through the computer network or fieldbus to accomplishthe control tasks. NCS has enormous advantages such as the information resources can beshared, less line connecting, easy to system extension and maintenance, higher reliabilityand ?exibility, and so etc. However, problems such as network-induced delay, packetsdropout are unavoidable during transmission because of the constraints of the communi-cation bandwidth, carrying capability and quality of service. The phase lag is inducedbecause of time delay, which can degrade the performance of control systems seriously,especially long time delay can even make control system unstable. The system can notcalculate using the newest sampled information or the plant can't be controlled in timebecause of packet dropout, which can damage message integrity, real-time property andalso degrade the performance of control systems. So the new controllers should be de-signed for the requirements of performance in the networked reconstruction of traditionalcontrol system.The objective of this paper is to present the systematic approaches to the designof the state-feedback controller and output-feedback controller for the networked controlsystem with long time delay or packet dropout, which can guarantee the performance ofclosed-loop system. And a real time simulation platform for networked control systemsbased on NS2 is also designed and realized. The main idea of these approaches is to modelthe NCS as Markov jump linear system (MJLS) and give the Linear Matrix Inequality(LMI) conditions on mean square stability and disturbance decay. Then new robustcontrol methods are developed.Considering the networked control system with long time delay, we present the designschemes of the mixed H2 /H∞state-feedback controller. A new model of the NCS withlong time delay is proposed using bu?ering method, the uncertain long time delay is notconverted to a maximum certain delay but to switch in some certain time delays, thusdecrease the delay increasement. The model has the form of the discrete-time MJLS. According to the theories of MJLS, the robust mean square stabilizing state controllers canbe constructed through minimizing an upper bound for the H2 -norm, under the restrictionthat the H∞-norm is less than a prespecified value for the two conditions that transitionprobability matrix is exactly known or is not exactly known but belongs to a convex set.The problem of the determination of the smallest H∞-norm is also addressed. We presentan approximate version of these problems via linear matrix inequality optimization. Thecontrollers have simple structure and can be realized easily.When some states of the plant can not be observed, switched output feedback con-trollers which can make the NCS stable and have the minimum H∞performance arederived using LMI method.Using the derived long time delay NCS model the closed NCS of output feedbacksystem is modelled as MJLS. Thus the delay increasement decreases. System performanceis measured via an H∞norm from disturbance to error and the controller is computed bythe necessary and su?cient LMI conditions according to the theories of MJLS. We cantranslate these conditions to LMI conditions using congruence transformation and Schurcomplement lemma which can be solved numerically.The data packets can be dropped during the transmission when packet collision andnetwork node failure occasionally occur. Considering the NCS with packet dropout, wethen study the problem of the stabilization and present the design schemes of H∞state-feedback controller and H∞output-feedback controller. The packet-dropout process isdefined as the sequence of time period between two successful data transmissions and isalso considered as the Markovian packet-dropout process. With a new freedom matrixintroduced, mean square stabilizing state feedback controllers can then be constructed.Furthermore, under the assumption that the disturbance is constant between the twosuccessful data transmissions, we model the NCS with data packet dropout as MJLS.Based on the results for MJLS, system performance is measured via an H∞norm and thecontroller is computed by a necessary and su?cient LMI. Then for the packet dropoutrate is known the NCS is modeled as discrete MJLS with two modes. According thebound real lemma of MJLS we can translate these conditions to LMI conditions whichcan be solved numerically using congruence transformation and Schur complement lemmato compensate the in?uence of packet dropout on system performance. Using these con-ditions a control engineer can determine the probability of packet dropout that can betolerated before the performance degradation is unacceptable.Simulations and experiments are necessary to invalidate the proposed H∞controlschemes. In this paper using network simulation software NS2 to simulate real networkperformance, we designed and realized real time simulation platform for networked con- trol systems based on NS2. The platform includes simulink controller/plant, Matlab datainterface, Java network interface and NS2 network simulation software. In our approach,we use Matlab, NS2 to simulate real time control and network performance separately.The communication between NS2 and Matlab is realized by Java application. Combinedthe virtues of NS2 and Matlab, and implemented real-time simulation for NCS. Finally asimulation example of networked system was given to establish that the simulation plat-form can simulate various network performances and also to analyze the various networkperformances such as time delay and packet dropout and the in?uence of network onsystem performance and stability.To invalidate the proposed H∞control schemes for the networked control systems,we discussed their applications in a practice problem using Matlab software and on thereal time simulation platform based on NS2 we have constructed: the networked invertedpendulum system. Set up the networked topological structure and choose the parametersin NS2 and then test the network performance times and again. After analyzing the datafrom the tests, we derive the time delay distribution and the packet dropout performance.The H∞controllers are designed according the proposed control schemes for the networkedinverted pendulum system. The simulation results indicate that the control e?ects withthe proposed approaches are valid and superior.In this paper NCS with long time delay and NCS with packet dropout are modeledas MJLS, the robust control schemes are proposed explicitly according to the theories ofMJLS and LMI method. The associated synthesis problems are formulated as optimiza-tion problems under LMI conditions, which can be solved in a numerical manner.The formulation processes and the proof of mentioned approaches are presented indetail in this thesis. Moreover, in order to validate the e?ciencies of the proposed ap-proaches, we give simulation results for each approach, which is discussed in detail fromparameters choosing, controller designing, network simulating. The simulation resultsindicate that the control e?ects with the proposed approaches are satisfactory.Deeper research work needs to be done since some problems are still remain to besolved, for example, the time delay considered in this thesis is assumes asτk∈[(l?1)h,lh)where l is an integer larger than 1. However in many cases the time delay induced bynetwork satisfyτk∈[0,lh) that make NCS very complex. In such cases how to modelthe NCS appropriately and robust control the NCS is a worthy research problem. Fur-thermore, we only discuss the NCS with time delay or packet dropout, how to design thestate feedback and the output feedback controllers for the NCS with both time delay andpacket dropout needs our further researches. For the real time simulation platform, thecontrolled plant is in virtual environment now that is imperfect for practical experiment and how to design and implement the semi-physical simulation platform needs our furtherresearches.