【作者】 敖志伟；

【导师】 过润秋；

【作者基本信息】 西安电子科技大学 ， 机械电子工程， 2006， 硕士

【摘要】 倒立摆是一个典型的快速、多变量、非线性、强耦合的自然不稳定系统。在控制过程中能反映控制理论中的许多关键问题,如镇定问题、非线性问题、鲁棒性问题以及跟踪问题等。对倒立摆系统的研究在理论上和工程应用上具有着深远的意义,相关的科研成果已经应用到航天科技和机器人学等诸多领域。本论文围绕一级、二级倒立摆系统,对包括模糊控制、神经网络、遗传算法在内的软计算以及它们之间的互相结合作了较为系统的讨论,研究了倒立摆系统的各种智能控制算法。对于一级倒立摆,利用神经网络的学习能力训练模糊控制器的隶属度函数,通过自适应神经模糊推理系统建立模糊控制器对倒立摆系统进行控制;对于二级倒立摆,运用最优控制方法设计融合函数减少模糊控制器的输入变量维数,成功解决了“规则爆炸”问题,利用专家知识设计Mamdani型模糊控制器的隶属度函数和模糊规则,并且利用遗传算法优化模糊控制器的参数,提升了模糊控制器的性能。最后通过编程实现了各种智能控制算法对倒立摆系统中的实物控制,均取得了令人满意的控制效果。控制结果表明,对不同的智能控制算法进行结合集成,能够综合吸取两种算法的优点,而将两者的缺点互相抵消。自适应神经模糊控制器的学习能力强,适用于时变对象;遗传算法作为一种启发式搜索算法,尽管学习时间比较长,但是遗传算法的全局搜索特性使其非常适用于模糊系统的设计与优化。更多还原

【Abstract】 Inverted pendulum is a typical fast, multivariable, non-linearity, strong- coupling and naturally unstable system. During its control process ,it can reflect many crucial questions in the control theory , such as calm question, non-linear problem, robust question as well as tracking question and so on.The research on inverted pendulum system has the profound significance in theory and project application.The correlative scientific research achievement has already applied to astronautics science technology and subject of robot and so many domains.This thesis encircled the inverted pendulum system, discusse the soft computing which including fuzzy control, nerve network(NN), genetic algorithms(GA) as well as their mutual combination systematically, study the the intelligent control algorithm of the inverted pendulum system. Towards to the single inverted pendulum , using the learning capability of nerve network to train membership function of the fuzzy controller, establishing a fuzzy controller to control the inverted pendulum through the Adaptive Neuro-Fuzzy Inference System(ANFIS).Towards to double inverted pendulums , it reduces the input variable dimension of the fuzzy controller by designing a fusion function using optimization control theory, solve the question of "rule explosion" successfully,and design the membership function and the fuzzy rule of Mamdani fuzzy controller using the expert knowledge,and optimizes the parameter of fuzzy controller using the genetic algorithms , promoted the performance of fuzzy controller .Finally realized the inverted pendulum system’s practicality control by each kind of intelligent control algorithm through programming , and obtained the satisfying control effect.The control result indicated that, the combination of two or more different intelligent control algorithm, can absorbs their merits, and counterbalances their defects mutually.Due to Adaptive Nerve-Fuzzy Controller has strong learning ability, it is suitable for time-variable object. As a kind of heuristic search algorithm, though its learning time is long,but its overall search characteristic enables it to apply in the design and optimization of fuzzy system.更多还原