A Study Of Networked Control Systems Based On Hidden Markov Models

Networked control system (NCS) is a kind of distributed real-time control system. The important feature of such NCS is that, instead of point-to-point connection, sensors, actuators and controllers are all connected to the network as nodes. The disadvantage inherent to NCS is the network induced delay and packet loss which occur while exchanging data through the shared medium. In order to provide the best performance, a network model is needed to describe the dynamic feature of network state. Hidden Markov Model (HMM) is used to model the networked control system, and a close-loop control pilot platform over campus network is built up. The main research work is as follows:NCS' background, basic concept, advantages and disadvantages are introduced. A survey of NCS research methods in control theory, network schedule and network devices is given.The concept, classification, basic problems and algorithms of HMM are introduced. Following that, the advantage of modeling the network with HMM is analyzed.A platform over campus Ethernet for experiment is designed. The hardware and software structures are introduced. The concepts and critical implements of main modules are given. With the platform, we get the QoS (Quality of Service) data of the network.A HMM is built for NCS with QoS data collected which includes both the time delay and packet loss information. To evaluated the model, A close-loop control experiment over campus network is designed. The experiment results show the validity of the hidden Markov model. Finally, a realtime HMM is given to improve the performance of the networked control system.The thesis concludes with a summary and perspectives of future research of networked control system.