【作者】 唐斌；

【导师】 刘国平； 桂卫华；

【作者基本信息】 中南大学 ， 控制科学与工程， 2008， 博士

【摘要】 网络化控制系统适应了控制系统逐渐地向分散化、网络化、智能化方向发展的趋势,从而成为当前控制领域的研究热点。本文针对网络化控制系统的反馈通道和控制通道中存在网络诱导时延和数据包丢失现象,研究了网络化控制系统的状态估计、LQG最优控制、广义预测控制和时滞相关镇定问题。主要内容如下:1)提出了基于预测思想的网络化状态估计算法,以补偿网络诱导时延和数据包丢失的影响。分析了该预测估计算法的性能,表明了虽然该算法具有明显的补偿效果,但是随着网络诱导时延或数据包丢失的增长,预测估计偏差略微递增,讨论了估计算法的稳定条件,通过仿真和实验验证了算法的有效性和分析的正确性,从而为网络化LQG最优控制和基于状态空间模型的网络化广义预测控制的实现提供了状态估计方法。2)结合网络化预测状态估计算法,提出了补偿网络诱导时延和数据包丢失的网络化LQG最优控制算法。基于动态规划的思想给出了补偿网络诱导时延和数据包丢失的LQ最优控制器的设计方法,讨论了通过网络传送的控制信号对状态估计影响,给出了网络诱导时延或数据包丢失同时存在于反馈通道和控制通道的情况下网络化预测估计算法,证明了分离定理的成立,通过仿真和实验验证了算法的有效性和分析的正确性。3)提出了采用最小预测步长补偿网络诱导时延和采用预测控制向量补偿数据包丢失的基于多项式模型的网络化广义预测控制算法,讨论了网络诱导时延和数据包丢失与最小预测步长和预测控制向量之间的关系,基于滚动优化方法给出了基于多项式模型的网络化广义预测控制器的设计方法。基于多个数据打包传送以补偿网络诱导时延和数据包丢失的网络化递推最小二乘辨识算法,进一步讨论了网络化预测自校正控制器的设计方法。通过仿真和实验验证了算法的有效性和分析的正确性。4)提出了采用最小预测步长补偿网络诱导时延和采用预测控制向量补偿数据包丢失的基于状态空间模型的网络化广义预测控制算法,给出了基于状态空间模型的网络化广义预测控制器的设计方法,通过仿真和实验验证了算法的有效性和分析的正确性。5)提出了一种改进的补偿网络诱导时延和数据包丢失的网络化时滞相关镇定方法。以线性标称系统为受控对象,该改进方法通过考虑过去在估算Lyapunov泛函导数时被忽略的一个积分项,来克服过去求导结果的保守性,给出了基于Lyapunov方法的指数渐近稳定的条件和相关结果,仿真验证了改进方法的优越性,并通过实验结果表明了方法的有效性。对于线性不确定系统,进一步研究了基于时滞相关方法的网络化保性能控制器的设计方法,给出了最优保性能控制器的求解算法,通过仿真和实验验证了算法的有效性和分析的正确性。更多还原

【Abstract】 Networked control systems have attracted much attention because they adapt to the distributed, networked and intelligentized development direction of control systems. This thesis is concerned with state estimation, LQG optimal control, generalized predictive control and delay-dependent stabilization of networked control systems, which takes network-induced time delay and packet loss in both the feedback channel and control channel into account. The main contents are given as follow:Firstly, a prediction-based networked state estimation algorithm is proposed to compensate for the influence of network-induced time delay and packet loss. The performance of the algorithm is theoretically analyzed, and it is pointed that although it achieves a good effectiveness on compensation, its estimation error appreciably increases with the increment of network-induced time delay and packet loss. The stability condition of this algorithm is also discussed. And simulation and experiment results are given to illustrate the effectiveness of algorithm and correctness of theoretical analysis.Secondly, using the prediction-based networked state estimate- ion algorithm, a networked LQG optimal control algorithm is proposed to compensate for the influence of network-induced time delay and packet loss on control performance. The design method of networked LQ optimal controller, which can compensate for the network-induced time delay and packet loss, is given using dynamical programming. With the analysis of the influence of control signal transported through network, the prediction-based networked state estimation is further developed under the condition that network-induced time delay and packet loss simultaneously exist in both the feedback channel and control channel. The separation theorem is proved to hold. Simulation and experimental results are given to demonstrate the effectiveness of algorithm and correctness of theoretical analysis.Thirdly, a polynomial-model-based networked generalized predictive control algorithm is proposed which compensates for the influence of network-induced time delay and packet loss using minimal predictive step and control increment vector, respectively. The relations between network-induced time delay and minimal predictive step, and between packet loss and control increment vector are discussed. With the iterative optimization approach, the design method of polynomial-model-based networked generalized predictive controllers is given. Furthermore, a networked generalized predictive self-tuning controller is discussed, based on the networked iterative least square identification algorithm which reduces the influence of network-induced time delay and packet loss with output prediction. Simulation and experiment results are given to show the effectiveness of algorithm and correctness of theoretical analysis.Fourthly, a state-space-model-based networked generalized predictive control algorithm is proposed in the similar way to the polynomial-model-based case. The design method of the controller is also given. Simulation and experiment results are given to show the effectiveness of algorithm and correctness of theoretical analysis.Fifthly, an improved method of networked delay-dependent stabilization, which can counteract the influence of network-induced time delay and packet loss, is proposed. With a linear nominal system as a controlled plant, the improved method overcomes the conservativeness of the derived condition by considering an integral item which was ignored in the past research of networked control systems. For an uncertain linear system, the design method of networked guaranteed cost controller is discussed with delay-dependent approach. Numerical and experiment results are given to illustrate the effectiveness of the algorithm and correctness of theoretical analysis.更多还原