【作者】 时维国；

【导师】 邵诚；

【作者基本信息】 大连理工大学 ， 控制理论与控制工程， 2012， 博士

【摘要】 网络控制系统是随着自动控制技术、网络通信技术、计算机信息技术的飞速发展而出现的一种新型的分布式、智能化、网络化实时反馈控制系统,以其交互性好、易于扩展和维护、可靠性高、资源共享等优点,在远程遥控操作、无线网络机器人、汽车、高速动车、兵器系统、航空航天领域,以及基于现场总线及以太网的工业控制系统等诸多领域具有广泛的应用前景。因此,对网络控制系统的相关问题进行研究是当前控制理论领域研究的热点,具有重要意义。网络的介入不可避免地在控制系统中引起网络诱导时延、数据包丢失等问题,使得传统控制系统的研究方法难以直接应用于网络控制系统,迫切需要研究现有的控制方法怎样才能应用于网络控制系统中,以及探索研究新的基于网络的控制理论与方法。同时,受网络带宽资源有限的限制,如何采用合适的调度方法合理分配网络资源对提高网络利用率以及控制系统的性能也至关重要。因此,网络控制系统的性能不仅取决于控制策略的优劣,同时还依赖于网络中信息调度策略的优劣,两者是相关联的、缺一不可的。为此,本文主要进行了如下几方面的研究：1.针对网络控制系统具有随机不确定时延的特点,通过对大量实测网络时延数据样本的分析,提出了网络时延的自适应AR预测模型。该模型在兼顾算法的预测精度和实时性的基础上,确定了模型的阶数,并利用最小均方算法根据新的时延值在线调整模型的加权系数,进而通过建立的模型对网络时延进行预测。仿真实验表明基于参数自校正AR模型的时延估计算法对时延有很好的预测效果。2.针对传统广义预测控制对时延补偿实时性差,提出了一种对网络长时延补偿的改进广义预测控制算法。该算法根据时延的估计值来确定合适的预测步长及控制时域,避免了求解Diophantine方程等复杂计算,降低在线计算量。仿真实验表明该方法对网络时延具有很好的补偿效果,有很好的实时性,对干扰和丢包具有良好的鲁棒性。同时提出了一种容易实现反馈控制的长时延状态观测器,实现了极点的任意配置,仿真实验表明该方法保证了系统响应的快速性和稳定性。3.针对变采样周期对系统控制性能的影响,分析确定了NCS采样周期的受限条件,提出了一种递推式变采样周期在线动态调度算法,可根据系统输出与均值的偏差、方差以及系统输出振荡情况来在线调节系统的采样周期,以绝对误差积分指标IAE等来衡量控制回路性能,仿真实验表明基于递推式变采样周期调度算法有效地改善了网络控制系统的控制性能,保证了系统的稳定性。4.针对网络控制系统中网络带宽有限等问题,考虑系统误差、误差变化率和网络利用率等因素,提出了一种模糊动态反馈变采样周期调度算法。该方法可根据系统误差及误差变化率运用模糊推理的方法来确定回路的优先级。通过网络利用率及控制性能指标来求取网络控制系统中各个回路的采样周期,合理地分配了网络有限带宽资源。仿真实验表明该方法提高了系统的控制性能,改善了网络的服务质量。更多还原

【Abstract】 With the rapid development of automatic control technology, network communication technology, computer technology and information technology, as a new type of distributed, intelligent, and network-based realtime feedback control system, Networked Control System (NCS) is used widely in remote control operation, wireless netwok robot, automotive, CRH train, certain weapons systems, aerospace, as well as fieldbus and industrial ethernet based industrial control systems and many other areas for the advantages such as low cost, easy to extend and maintain, high efficient, flexible, and resource sharing, and the abroad application foreground. Therefore, some research issues related to NCS are of great significant in the current field of control theory.The traditional control methods are difficult to be applied directly to NCS for network-induced delay, packet loss and other issues caused by the network intervention in the control system. There is an urgent need to study how to extend the existing control methods to NCS, and explore new control theory and methods based on network. Subject to network bandwidth limitation, at the same time, it is important to allocate reasonably network resource through available scheduling methods in order to improve the utilization of network and the performance of control system. Therefore, the performance of NCS depends not only on the merits of the control strategy, but also on the effectiveness of network scheduling policy, which need to be considered together. Therefore, this dissertation is made and main contributions are concluded as follows:Firstly, according to the random uncertain time-delay of networked control system, a large number of measured delay sample data is analyzed and treated, then the adaptive AR predictive model is proposed. On the basis of consideration on prediction precision and real-time property in algorithm, the model ascertains its order, and uses the least mean square algorithm based on the new time-delay in order to adjust the weighting factor online. The simulation results show that the delay estimation algorithm based on self-tuning AR model has much better prediction effect on delay.Secondly, for traditional GPC used to compensate delay is short of real-time property, an improved GPC algorithm is proposed to compensate random long delay of network. The predictive step and control horizon of GPC are determined according to the estimated value of network time delay. The method avoids solving the Diophantine equations in order to reduce the computational quantity. The simulation results show that the method has very good effect on delay compensation, and well real-time property, and good robustness on interference and loss package. A state observer of long time delay is put forward simultaneously, which can realize easily state-feedback control and arbitrary pole configuration. The simulation results show that the method has good effect on delay compensation and good stability.Thirdly, the constraints of sampling period is analyzed and determined, and an online dynamic scheduling algorithm with recursive and variable sampling period is put forward, considering the impact of variable sampling period to network control performance. It can adjust the sampling period according to the mean, variance and oscillation of system output. It reflects the QoC of NCS through error absolute integral index IAE. The simulation results show that the scheduling algorithm effectively improve the control performance and the stability of system.Finally, subject to network bandwidth limitation in NCS, a fuzzy dynamic feedback scheduling algorithm of varying sampling period is proposed, by considering the error, error rate and network utilization. It uses the fuzzy inference method to determine the priorities of loops according to the error and error rate, and uses the network utilization and control performance index to determine the sampling period of each loop in NCS, so it distributes reasonably network limited bandwidth resources. The simulation results show that the method can improve the performance of NCS, and improve the quality of network service.更多还原