【作者】 王海涛；

【导师】 王洪瑞； 宋维公；

【作者基本信息】 燕山大学 ， 控制理论与控制工程， 2003， 硕士

【摘要】 机器人运动控制的智能控制方法主要包括模糊控制、神经网络控制、变结构控制以及相互融合技术。这些智能控制方法在多自由度刚性机器人轨迹跟踪方面的应用是十分有效的，在实际应用中操作性能也是相当优秀的。在实际工程中，要得到机器人精确的数学模型是一件很困难的事情，因此，我们在建立机器人数学模型时，需要做合理的近似处理，而忽略一些不确定性因素，诸如参数误差、未建模动态、观测噪声和不确定性的外界干扰等等。然而这些不确定性的存在可能会引起控制系统品质的恶化，甚至成为系统不稳定的原因。因此，针对机器人轨迹跟踪，本文引入智能控制方法，使机器人满足高速高精度的要求。本论文首先介绍了适合不确定机器人实时控制需要的小脑模型联结控制器(Cerebella Model Articulation Controller)，简称CMAC。用CMAC网络来逼近机器人的不确定性函数。由于CMAC的泛化能力与存储容量之间的矛盾是不可忽视的，因此本文提出一种模糊CMAC方案解决了这个矛盾，并保证了系统的快速稳定性。其次，变结构控制因其不需要被控对象精确的数学模型和对参数变化以及噪声干扰的不敏感性，所以尤其适合于控制不确定的机器人系统，但是其控制的鲁棒性与抖振又是并存的，所以削弱抖振，并保证系统的稳定性是十分必要的。本文提出一种基于系统状态的模糊变结构控制方法，有效地削弱了抖振，同时又保证了机器人的轨迹跟踪精度。最后，考虑对机器人的参数不确定性和外界干扰分别进行补偿的控制策略，用神经网络智能控制方法逼近机器人参数不确定性函数，对于外界干扰和逼近误差，由于它们会使系统状态长时间到达不了稳定状态，为此本文分别采用两种控制方案代替传统的变结构控制方法，达到消除逼近误差，抑制外界干扰的目的。通过分析所设计控制器的稳定性、鲁棒性等品质指标，并且对同一个二自由度串联机器人计算机仿真，证明了本论文所提出控制方案的有效性和可行性。更多还原

【Abstract】 The robotic kinetic intelligent control methods mainly include fuzzy control, neural network control variable structure control and their coalescent control. These intelligent control methods are very effective for the multiple- degrees-of-freedom rigid robotic manipulators trajectory tracking, and make the robot have quite excellent practical performance in the application fields. In practice, unfortunately, it is impossible to obtain a perfect or even reasonably accurate dynamics model of a robotic manipulator. Therefore, when we establish robotic model, we need make reasonably approximate treatment, and ignore some uncertain factors, for example, parameter errors, unmodeled dynamics, observed noises and uncertainly external disturbances and so on. The uncertainties probably result in deterioration of control system quality, even turn in instability factors of robotic system. In this dissertation, the intelligent control schemes are introduced in the control system to eliminate the influence of the uncertainties.This dissertation first introduce the Cerebella Model Articulation Controller, which is satisfies real-time control requirement for uncertain robotic manipulator. The CMAC network is used to approximate robotic uncertain function. Because the contradiction between the generalization ability of CMAC and the memory capacity is not be neglected, a fuzzy CMAC strategy is presented to solve the contradiction, and the system stability is guaranteed as well. Second, variable structure control method need not accuracy mathematic model of controlled device and is insensitivity to the parametic variation and noise disturbance, so it is specially suit to the robotic control. But its robustness and control chatting are associated. Weaken chatting and guarantee system stability is necessary. A fuzzy variable structure control strategy based on system state is proposed to lower chatting effectively and guarantee the trajectory tracking accuracy of robot. Last, the robotic parameter uncertainties and external disturbances <WP=6>are separated to be compensated. The neural network intelligent controller is utilized to compensate for the former. The external disturbances and approximation errors will delay the system states reaching stability condition. This paper presents two control strategies instead of conventional variable structure control. The presented controllers can cancel approximation errors and restrain disturbances.The performances of controller such as stability and robustness are analyzed. The simulation results are presented for the same 2-DOF serial robotic manipulator, which validate the effectiveness and feasibility of the proposed schemes.更多还原