【作者】 崔根群；

【导师】 张明路；

【作者基本信息】 河北工业大学 ， 机械制造及其自动化， 2007， 博士

【摘要】 由于化学危险品泄漏的危害性,各类化学反应容器和输送管道的泄漏检测与维修已经成为石化工业安全生产、预防重大事故发生所关注的问题。在危险环境中,具有自主能力的移动机械手便成为代替和辅助人类完成一定任务的最佳选择。本文在国家863计划项目“化学危险反应器泄漏检测与修补移动机械手系统”(项目编号:2003AA421040)的支持下,针对所研制的P3-AT型移动机械手进行了运动学、动力学分析,采用人工神经网络研究了移动机械手的运动控制技术。主要的研究内容与创新如下:1、描述了所设计的P3-AT型移动机械手的基本结构特征,分析了五自由度机械手和两自由度轮式移动载体(平台)的运动特点。分别建立了移动载体子系统、机械手子系统以及移动机械手系统整体的运动学模型,进行了正、逆运动学分析。在ADAMS分析软件中建立了机械手的虚拟样机,分析了机械手的工作空间和特殊工作位置,得到了机械手实际应用中的有效工作区域。2、根据移动机械手系统的复杂性,提出了运动规划策略,构建了简化的移动机械手系统结构模型,利用牛顿——欧拉方法推导了简化移动机械手系统的动力学模型。通过仿真分析了移动平台与机械手间的动力耦合作用,仿真表明移动平台的加速度值愈大,其对机械手关节负载力矩的影响愈大,并影响移动机械手的轨迹跟踪和定位精度。建立了移动平台子系统、机械手子系统的动力学理论模型,并对各子系统的动力学进行仿真分析,得到了机械手各关节转矩与位姿的关系和机械手几个较危险位姿;对于移动平台,通过控制其左右前轮的驱动力矩,可以实现移动平台按给定的轨迹运动,同时也能实现移动平台方向角的改变。3、提出了基于神经网络的移动机械手的分层递阶智能控制策略,采用三层控制实现移动机械手的部分或完全自主控制。决策层进行任务规划;处理层包括两个径向基函数(RBF)神经网络子控制器,分别对移动载体和机械手的动力学进行控制和补偿;执行层依据处理层输出的信息来独立控制各驱动电机,实现所需的运动。RBF神经网络子控制器应用李雅普诺夫稳定性设计方法,通过分别建立移动载体、机械手和移动机械手系统的李雅普诺夫方程,由第二类李雅普诺夫方法建立了移动载体和机械手的动力学补偿,并应用Matlab对所建立的RBF网络进行训练仿真,仿真结果表明了RBF神经网络的有效性和可靠性。4、针对移动机械手硬件体系结构进行了设计,其中包括移动载体、五自由度机械手,视觉CCD传感器和超声传感器模块等。介绍了移动机械手的控制软件设计,包括PMAC上运行的运动程序和移动载体PC上运行的主控程序。应用Matlab编程软件对五自由度机械手的运动及控制进行了仿真,验证了机械手操作的运动准确性。并在实验环境下对移动机械手系统进行了实验,验证了移动机械手运动的准确性及控制的有效性和可靠性。更多还原

【Abstract】 Due to the harmfulness of deleterious chemistry leakage, the measurement and maintenance of chemistry reactors and carrying pipeline has arosed the attention of researchers. The mobility and manipulation capability of a mobile manipulator provides convenience to accomplish some tasks in dangerous environment. This dissertation is carried on the research to the dynamics and control technology for the mobile manipulator designed, supported by the China National Hi-tech R&D Program (863 Program)—Study on the mobile manipulator for measurement and maintenance of chemistry reactors (Granted No.2003AA421040).1. The structure of P3-AT mobile manipulator designed is described. The kinematics character of 2 DOF mobile platform and 5 DOF manipulator is analyzed. The kinematics modeling of mobile platform and manipulator is studied respectively, then the kinematics modeling of mobile manipulator system is derived. And the forward and inverse kinematics equations are derived. Based on the research above, the simulation model of 5 DOF manipulator is developed in the software ADAMS, which can be used to analyze motion space and dynamics visually.2. The strategy of coordination-motion planning for the mobile manipulator is presented, and a simple mechanism system model of the mobile manipulator is built. Then the dynamic modeling of the system is studied through the method of Newton-Euler. By dynamic computer simulation, the interaction between mobile platform and manipulator is analyzed. The analysis result indicated that the larger is the acceleration of mobile platform, the larger the force of manipulator applied by the platform is. It is harmful effects for position accuracy of mobile manipulator. And the theoretical dynamic models of them are derived by the method of Newton-Eular and Lagrange separately. By the computer simulation, dynamic characteristic of two subsystem is analyzed. The relation between joint moment and posture of manipulator is displayed, and the bad postures of manipulator are obtained. The mobile platform may achieve motion trace and alter azimuth as demand.3. Based on the kinematic and dynamic motioned above, a hierarchical intelligent controller based on neural network is proposed for the coordinated control of the mobile manipulator, which mimics human behavior. It consists of three levels: decision-making level, processing level and execution level. Decision-making level is a task-planning unit. Processing level includes two RBF neural network controllers, in which unknown mobile platform and manipulator dynamic parameters are identified and compensated. Execution level controls the movement of each motor of mobile manipulator, based on the output control torques from processing level. In RBF neural network controllers, generalized Lyapunov equations of two subsystem (mobile platform and manipulator) and mobile manipulator system are derived, using Lyapunov stability theory. Then the simulation of sample training in RBF neural network is completed by Matlab software. The result shows the training in RBF neural network is effective and reliable.4. The hardware system of the mobile manipulator is designed. It includes a mobile platform and a 5 DOF manipulator, CCD sensors and ultrasonic sensors module, etc. The control program of mobile manipulator is developed, including movement program in PMAC and central control program in PC of the mobile platform. Then the motion and control simulation of manipulator is completed by Matlab software, which displays the action accuracy of manipulator. In the end, a experiment of mobile manipulator is accomplished. The experiment and simulation results show that kinetic control of manipulator is precise and reliable.更多还原