Adaptive Identification and Control of Dynamical Systems Using Neural Networks

本文的研究重点是基于神经网络的复杂非线性动态系统的自适应辨识与控制。从理论的角度来看，其主要目的是产生稳定的自适应规律来调节系统的参数。从实际的角度来看，目标是生成在复杂环境中产生高性能的控制方法。神经网络的非线性特性使得整个系统的稳定性分析成为一项极其困难的任务。因此，有必要把我们的注意力集中在至少在经验上表现成功的一类系统上。因此，研究将分以下四个阶段进行：(i)为识别和控制模型确定合适的动态结构，这些模型适用于广泛的非线性植物；（ii）开发动态反向传播方法来调整控制器参数；（iii）使用这些方法对复杂系统的识别和控制进行广泛的计算机模拟；（iv）对在（iii）中证明有效的系统类别进行稳定性分析。

The focus of this research is on the adaptive identification and control of complex nonlinear dynamical systems using neural networks. From the theoretical point of view the main aim is to generate stable adaptive laws for the adjustment of the parameters of the system. From the practical viewpoint the aim is to generate control methods which yield high performance in complex environments. The nonlinear nature of the neural networks makes stability analysis of the overall system and extremely difficult task. Hence, it is essential to confine our attention to a class of systems which have been shown at least empirically to perform successfully. The research will consequently be carried out in the following four stages: (i) determination of suitable dynamic structures for identification and control models which are applicable to a wide class of nonlinear plants, (ii) development of dynamic back propagation methods for adjusting controller parameters, (iii) use of such methods to carry out extensive computer simulations on the identification and control of complex systems (iv) stability analysis of the class of systems which prove efficient in (iii).