网络控制系统随机通信时延的分析与补偿

发布时间：2018-04-19 16:13

本文选题：网络控制系统 + 随机通信时延　；参考：《江南大学》2017年硕士论文

【摘要】：近年来,随着控制、网络和计算机等技术的不断发展,控制系统发生重大变革,其不仅要求系统内部实现完全的分散控制,同时要求与Internet和Web等通用网络连接,网络控制系统(Networked Control Systems,NCSs)应运而生。有限的网络带宽导致不可避免的随机通信时延和数据丢包,使得系统丧失定常性、因果性、完整性和确定性,越来越多的国内外学者致力于设计稳定、可靠、高性能的网络控制系统。网络通信时延作为研究的难点,对其进行特性分析和补偿有着重要的理论意义和广泛的应用价值,也是本课题的研究方向,主要的研究内容如下:针对网络控制系统中双向通道的随机时延问题,提出一种基于广义预测控制的时延补偿设计方法。闭环系统的时延补偿控制器由网络预测器和时延补偿器构成,分别用于补偿具有不同特性的反馈通道时延和前向通道时延,给出一种依据随机时延选取控制参数的方法以提高系统性能,并得到了闭环控制系统的稳定条件。最后采用TrueTime工具箱仿真实际的网络通信,对固定时延和随机时延环境下的补偿性能分别进行分析,并讨论控制参数取值对系统性能的影响。仿真结果表明,时延补偿控制器对固定时延和随机时延都有效,而依据时延选取控制参数的系统相比固定参数的系统具有更好的性能。针对网络控制系统中前向通道随机时延带来的不确定性问题,在时序分析的基础上,提出一种基于自回归模型(Auto Regressive,AR)的时延在线多步预测模型。反馈通道发送端将执行器端的时延信息与传感器端的采样信息一并打包发送,远程控制器事件触发接收数据包,取出时延信息更新时延对应的时间序列,经数据预处理后用于建立自回归模型,并由递推最小二乘法在线更新模型参数。最后,通过滚动预测实现前向通道时延的多步预测,获得当前时刻的前向通道时延预测值。基于Matlab/Simulink平台,采用TrueTime工具箱构建网络闭环仿真系统,实验结果表明,与基于广义回归神经网络的在线预测模型相比,所提出的在线多步预测模型具有更好的预测性能。基于前向通道时延的在线多步预测模型,提出一种改进的网络控制系统随机时延补偿设计方法。采用区间分割对通信时延的随机性进行描述,进一步降低时延的预测误差,获得可信的环回时延预估值。针对环回时延设计补偿控制器,主要由状态预估器、网络预测器和时延补偿器构成,状态预估器实现前向通道时延的预测并对受控对象进行状态预估,将获得的时延预测值、控制输入与输出预估值供给网络预测器以获得优化控制增量序列,同时给出了控制参数的设置方法。采用TrueTime工具箱构建网络控制仿真系统,结果表明,基于时延预测模型改进的时延补偿控制器对随机时延有效,相比传统的补偿控制方法具有更好的控制性能。
[Abstract]:In recent years, with the continuous development of control, network and computer technology, the control system has undergone major changes. It not only requires complete decentralized control within the system, but also requires connecting with Internet and Web.Networked Control Systems (NCSs) emerged as the times require.Limited network bandwidth leads to inevitable random communication delay and data packet loss, which makes the system lose its stability, causality, integrity and certainty. More and more scholars at home and abroad are devoted to design stability and reliability.High performance network control system.Network communication delay is a difficult point in the research. It has important theoretical significance and wide application value to analyze and compensate its characteristics. It is also the research direction of this topic.The main research contents are as follows: aiming at the problem of random delay of two-way channel in networked control system, a design method of delay compensation based on generalized predictive control is proposed.The time-delay compensation controller of the closed-loop system consists of a network predictor and a time-delay compensator, which is used to compensate the feedback channel delay and the forward channel delay with different characteristics, respectively.A method of selecting control parameters based on random delay is presented to improve the performance of the system and the stability conditions of the closed-loop control system are obtained.Finally, the TrueTime toolbox is used to simulate the actual network communication, and the compensation performance in the environment of fixed delay and random delay is analyzed, and the influence of control parameters on the performance of the system is discussed.The simulation results show that the time-delay compensation controller is effective for both the fixed delay and the random delay, and the system with the control parameters selected according to the delay has better performance than the system with the fixed parameters.In order to solve the uncertainty problem caused by the stochastic delay of forward channel in networked control system, an online multistep prediction model of time delay based on autoregressive model (AR) is proposed based on time series analysis.The feedback channel sender packages the delay information of the actuator side and the sampling information of the sensor side together. The remote controller event triggers the receiving data packet, and the delay information updates the time series corresponding to the delay.The autoregressive model is established after data preprocessing, and the model parameters are updated online by recursive least square method.Finally, the multi-step prediction of forward channel delay is realized by rolling prediction, and the predictive value of forward channel delay at current time is obtained.Based on Matlab/Simulink platform and TrueTime toolbox, the network closed-loop simulation system is constructed. The experimental results show that the proposed on-line multi-step prediction model has better prediction performance than the on-line prediction model based on generalized regression neural network.Based on the online multistep prediction model of forward channel delay, an improved design method of stochastic delay compensation for networked control systems is proposed.The randomness of communication delay is described by interval segmentation, which can further reduce the prediction error of delay and obtain reliable loop back delay preestimate.A compensation controller is designed for loop time delay, which is composed of state predictor, network predictor and time delay compensator. The state predictor can predict the time delay of the forward channel and predict the state of the controlled object.The control input and output prevaluation is supplied to the network predictor to obtain the optimal control increment sequence, and the method of setting the control parameters is given.The simulation system of network control is constructed by using TrueTime toolbox. The results show that the improved delay compensation controller based on delay prediction model is effective for stochastic delay and has better control performance than the traditional compensation control method.
【学位授予单位】：江南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP273

【参考文献】