【作者】 么健石；

【导师】 徐心和；

【作者基本信息】 东北大学 ， 控制理论与控制工程， 2006， 博士

【摘要】 图式理论(Schema Theory)作为分析人类等生物复杂动态系统的方法论，起源于哲学并发展于心理学，70年代后期被应用到人工智能领域，同时受到控制界的极大关注，已经成为智能控制的一个新的研究方向。图式理论建构了人脑中有组织的知识结构的框架，这个框架是人对某一范畴事物典型特征及关系的抽象，包含了客观环境和事件的一般信息。图式理论的可变性、结构性和主动性特征，使其在宏观上建立了认知科学与脑理论沟通的桥梁，在微观上实现了人工智能功能模拟和结构模拟的统一，并在机器人控制、神经网络系统和遗传算法等领域应用中取得了丰硕成就。 本文将图式理论应用于一类欠驱动机械系统的控制研究。欠驱动机械系统(Underactuated Mechanical Systems)是一类构成系统的广义坐标维数多于控制输入维数的非线性系统。例如倒立摆系统、肩部受控的二连杆机械臂(Pendubot)、肘部受控的二连杆机械臂(Acrobot)和三连杆单杠体操机器人均属于这一类系统，其控制目标是将一级或多级摆杆从自然悬垂位置摆起到倒立点位置并使其稳定不倒。 论文主要研究内容： (1)欠驱动机械系统的动态特性分析。欠驱动机械系统是一个多变量、强耦合、自然不稳定系统，这类系统的共同特点是具有零动态不稳定，非完备以及反馈的不完全线性化等特性。通过拉格朗日动态分析对系统进行了分类并建立了标准型，即将系统分解成全驱动和无驱动两个子系统，实现了高阶系统的结构分解。 (2)详细综述了图式的概念、图式理论的形成发展及其在机器人控制领域中的应用，着重研究了李祖枢教授提出的感知—运动图式的建构理论和Jose M C提出的动态图式层级。感知—运动图式理论能够建立一类欠驱动机械系统中各子系统和线性化系统的图式集，实现控制过程从状态空间到任务空间再到动作空间的映射；动态图式层级则能够针对目标的控制流程，实现复杂非线性系统的分层递阶控制。本文以上述理论作为研究工作的理论基础，并对两种理论进行融合，建立了一类欠驱动机械系统的智能控制器。 (3)基于拉格朗日方法的动力学建模。由于欠驱动机械系统的状态变量比较多，因此其动力学模型比较复杂。基于系统齐次坐标变换和系统能量的变化，采用拉格朗日法分别建立各种具体欠驱动机械系统的动力学模型。更多还原

【Abstract】 The schema theory is regarded as the analyzing methodology for such complex dynamic systems of living beings as the mankind, etc. Originated from philosophy and developed in combination with psychology, it was applied into the artificial intelligence field in the late 1970’s. Arousing great concerns, it has already become a new research direction of intellectual control. The schema theory establishes a frame of the organized knowledge structure in the human brain, abstracting typical characteristics and relations of things in a certain category. Both objective environment and general information of the incident are included. Characterized as changeable, structural and initiative, the schema theory, on the macroscopic, bridges the cognitive science with the brain theory, on the microcosm unifies the function and structure simulations of artificial intelligence. Consequently it has made significant achievements in robot controlling, neural network system and genetic algorithm, etc.The schema theory is applied here controlling a class of underactuated mechanical systems, which is a kind of non-linear systems with fewer inputs than degrees of freedom. Typical examples include the inverted pendulum system, the shoulder controlled two-link robot arm system (Pendubot), the elbow controlled two-link robot arm system (Acrobot) and three-link gymnastic robot. The controlling goal is to swing up the single or multi-level pendulum from its natural hanging position to the inverted position and keep it stable there.The main research contents of the dissertation are summarized as follows:(1) Dynamic characteristic analysis. The underactuated mechanical systems are naturally unstable and multivariable systems with strong coupling. They are characterized by dynamic unstable zero, nonholonomic and incomplete feedback linearization. Through Lagrangian dynamic analysis, system classification is carried out and the normal form set up. The structure is decomposed by breaking up the high order system into a fully actuated subsystem and an unactuated subsystem.(2) Survey of the schema concept and the schema theory, as well as their applications in robot control fields, with emphasis on Professor Zushu Li’s perception-motor schema theory & Mr. Jose M C’s dynamic schema hierarchy. Perception-motor schema theory can establish更多还原