STUDY ON RECONSTRUCTION OF INTELLIGENT CONTROL SYSTEMS BY UTIUZING NONLINEAR H-INFINITY CONTROL AND COPUTATIONAL STATISTICS

1.导出了一类新的自适应控制器，它们对某些有意义的代价泛函是最优或次优的。针对一般自适应控制问题，构造了H ~ 2或H_∞最优（或次优）控制系统。2.推广了文献[1]的研究结果，针对非线性时变过程，提出了非线性自适应H_∞控制方案。将系统参数的时变元素和整定参数的估计误差视为外部干扰，得到了一类H_∞控制问题的解。3.推广了文献[2]的研究结果，针对含三层神经网络的非线性参数模型，提出了非线性自适应H_∞控制方案。所得到的控制策略是作为一类H ∞控制问题的解决方案而导出的，其中非线性系统的估计结构中的近似误差和算法误差 ...更多信息 4.针对线性变参数（LPV）系统，提出了一种自适应增益调度H ∞控制策略。在所提出的控制方案中，未知的预定参数的估计得到递归，这些被馈送到控制器，以稳定对象，并获得H ∞控制性能自适应。基于“有界真实的引理”中的线性矩阵不等式（LMI），利用李雅普诺夫方法进行了稳定性分析。5.将文献[2]和[3]的研究结果应用到[4]中，针对LPV系统，提出了自适应增益调度H ∞控制方案的鲁棒控制方案。稳定控制信号被添加到调节时变调度参数的效果，并且这些被导出为某些H ∞控制问题的解决方案。同样的控制方案被应用于具有非线性参数模型和时滞元件的LPV系统的增益调度控制。6.利用混合自适应律的迭代学习控制方案被开发用于机器人的运动控制。提出了梯度和最小二乘混合自适应律，并对整个系统进行了稳定性分析。此外，通过扩展混合自适应方案，提出了二维自适应控制程序。二维自适应控制结构同时包含离线和在线两种自适应过程，并提供了更巧妙的学习特性，自适应过程本身得到了自适应的改进。少

1.New class of adaptive controllers which are optimal or sub-optimal to certain meaningful cost functionals, were derived. The adaptive H2 or H-infinity optimal (or sub-optimal) control systems are constructed for genoral daises of adaptive control problems.2.By extending the study result of 1, the nonlinear adaptive H-infinity control schemes are developed for nonlinear time-varying processes. The resulting control strategy is derived as a solution for certain class of H-infinity control problems, where estimation errors of tuning parameters and time-varying elements of system parameters are regarded as external disturbances.3.By extending the study result of 2, the nonlinear adaptive H-infinity control schemes are developed for nonlinear parametric models including three-layered neural networks. The resulting control strategy is derived as a solution for certain class of H-infinity control problems, where approximation errors and algorithmic errors in the estimation structures of non … More linear parametric models, are reqarded as external disturbances.4.The adaptive gain-scheduled H-infinity control strategy is developed for linear parameter-varying (LPV) systems. In the proposed control schemes, estimates of unknown scheduled parameters are obtained recursively, and those are fed to the controllers to stabilize plants and attain H-infinity control performance adaptively. Stability analysis is carried out via Lyapunov approaches based on linear matrix inequalities (LMI) in "Bounded Real Lemma".5.By applying the study results of 2 and 3 into 4, the robust control version of the adaptive gain-scheduled H-infinity control scheme is developed for LPV systems. Stabilizing control signals are added to regulate the effect of time-varying scheduling parameters, and those are derived as a solution of certain H-infinity control problems. The same control schemes are applied to the gain-scheduled control for LPV systems with nonlinear parametric models and time-delayed elements.6.The iterative learning control schemes by utilizing hybrid adaptation laws are developed for motion control of robotic manipulators. The gradient and least squares hybrid adaptation laws are proposed, and stability analysis of overall systems is carried out. Additionally, by extending hybrid adaptation schemes, two-dimensional adaptive control procedures are proposed. The two-dimensional adaptice control structures contain 2 types of adaptation processes, off-line tuning and on-line tuning, simultaneously, and provide more skillful learning properties where adaptive processes themselves are improved adaptively. Less