作者： 翟宏昌
学号： 01200402060522
所属院系： 信息与控制工程学院
专业： 控制理论与控制工程
导师姓名： 李平
毕业时间： 2007-6-15
学位类别： 学历硕士
保密级别： 公开
关键词： 变论域，模糊控制，模糊RBF网络，倒立摆，OCD优化工具箱
英文关键词： Variable universe, Fuzzy control, Fuzzy RBF network, Double inverted pendulum, OCD optimal toolbox
中文文摘： 倒立摆是一个不稳定、高阶次、多变量、强耦合的非线性系统而被广泛研究，在自动控制领域中, 倒立摆仿真或实物控制实验,已成为检验一个新的控制理论是否有效的重要手段，同时也是验证一个新的控制方法的基础实验平台。许多控制概念如控制系统的稳定性、可控性、系统收敛速度和系统的抗干扰能力等，都可以通过倒立摆系统直观地表现出来。 倒立摆的研究一般可以分为两个方面: 设计控制器使倒立摆系统稳定于特定位置；倒立摆系统自动起摆然后立于指定位置。目前，倒立摆的研究主要集中在前者。 针对模糊控制在多变量控制时的规则组合爆炸问题，运用了状态变量合成方法，同时针对系统容易产生稳态误差，运用了变论域的思想。由此，提出了一种基于变论域的模糊RBF网络控制器的设计方法，分别对一级倒立摆和二级倒立摆和三级倒立摆进行控制。同时引入了OCD优化工具箱对控制器参数进行优化，使得控制效果更好。 对于一级倒立摆，采用了模糊控制、LQR控制、变论域的状态变量合成法以及模糊神经网络的控制方法。对于二级倒立摆，采用了变论域模糊RBF网络控制方法，并与LQR及变论域的状态变量的模糊控制进行了比较，控制效果更好。对于三级倒立摆，采用了状态变量合成方法。
英文文摘： It is widely accepted that inverted pendulum is studied for its unstable、 high-order、multi-variable、tight coupling、nonlinear system. In automation field, the simulation and practical control experiment has been the most important way of testing a new theory whether it is effective. At the same time, it has been the basic experimental panel on which a new control way is based. Inverted pendulum can show many characters of control system, such as stability、controllable、the speed of constringency and the ability of anti-jamming. Aimed at the fuzzy control rule number explosion problem in the multi-varieties fuzzy control, state variable synthesis method is put forward. At the same time, aimed at steady state error of the control system, varied universe thinking is carried out. Therefore, a control algorithm based on varied universe fuzzy RBF neural network is presented to control the single, double inverted pendulum and triple inverted pendulum. Furthermore, OCD optimal toolbox is employed to optimize the parameters of the controller in order to acquire the better control results. For single inverted pendulum, fuzzy control, LQR control, variable universe state variable synthesis method, fuzzy-neural network methods are used to control it. A variable universe fuzzy RBF neural network is used to control the double inverted pendulum. Compared with LQR control method and variable universe fuzzy control method, it has better control results. State variable synthesis method is used to carry out the control of triple inverted pendulum.
