基于视觉伺服的机械臂系统及控制方法研究

发布时间：2018-02-27 20:34

本文关键词： 视觉反馈机器人操作系统机械臂控制算法 RBF神经网络　出处：《中国矿业大学》2017年硕士论文　论文类型：学位论文

【摘要】：机械臂作为机器人家族中比较重要的一员,在工业和家庭服务等各个领域都有着越来越广泛的应用,因此以工业和服务机器人为基础的各项技术研究包括其传感方式、控制算法等也越来越受到重视。本文主要以目前的研究热点视觉伺服和机器人操作系统为基础,针对多关节机械臂在实现智能化过程中的基于视觉反馈的抓取、机械臂的操作系统和控制算法等问题,设计了基于视觉伺服的机械臂操作平台并对其控制算法进行了改进研究,从而帮助提高了机械臂的控制性能和智能化程度并提升了机械臂与人的协作能力。对机械臂进行了具体如下的研究:(1)本文先对机械臂的硬件平台进行设计和搭建:确定了以六关节轻型机械臂作为研究对象;然后本文为所选定的机械臂设计绳驱动柔性三指手作为机械臂进行抓取操作的执行端并为其设计了独立的控制方法。完整的硬件平台设计为之后进行的机械臂基于视觉伺服控制的研究提供了有力的保障。(2)机械臂控制系统由机械臂和视觉系统组成,是控制相对复杂的系统,因此其软件平台的建立至关重要。视觉伺服系统需要良好的视觉传感器,经过对比后我们选择了深度传感器Kinect并利用计算机视觉库对其所采集的图像进行处理,从而得到机械臂所要抓取物体的在机械臂工作空间中的位置信息;对于机械臂的控制我们在开源机器人操作系统框架下对机械臂进行操作和规划,利用机器人运动规划模块为机械臂添加控制包,并在可视化界面中对其进行监视和规划。(3)同时为了满足机械臂的智能控制,本文对机械臂进行运动学分析,包括正向运动学和逆向运动学并求得其雅克比矩阵。并且再此基础上通过对机械臂动力学方程的推导,将面向工作空间中的RBF神经网络算入引入机械臂的控制,并用Matlab的Simulink仿真模块对该算法在机械臂中的应用进行仿真并对RBF控制算法的仿真结果进行分析。(4)最后通过介绍本文所搭建的系统的视觉反馈系统的具体工作方式,对机械手模型进行URDF转换使其可以通过Rviz显示其状态以利于我们在机械手抓取时对其状态的检测。最后,使用本文所搭建的机械臂系统进行抓取、精度测量试验以及画线试验,证明了基于视觉反馈的机械臂控制系统可以控制机械臂准确稳定的抓取物体并验证了 RBF神经网络控制算法可以提高机械臂的控制精度和控制的稳定性。本论文从硬件、软件到控制算法操作平台等方面完整的建立了机械臂的研究体系,为机械臂之后的研究和进一步智能化打下了坚实的基础。该论文有图79幅,表7个,参考文献102篇。
[Abstract]:As an important member of robot family, manipulator has been widely used in many fields, such as industry and home service, so the research of technology based on industry and service robot includes its sensing method. The control algorithms are paid more and more attention. This paper is mainly based on the current research focus of visual servo and robot operating system, aiming at the multi-joint manipulator in the realization of intelligent process based on visual feedback grab, Based on the operating system and control algorithm of manipulator, a manipulator operating platform based on visual servo is designed and its control algorithm is improved. Thus, it helps to improve the control performance and intelligence of the manipulator and enhance the cooperation between the manipulator and the human. The specific research on the manipulator is as follows: 1) the hardware platform of the manipulator is designed and built in this paper. The light mechanical arm with six joints is chosen as the research object. Then this paper designs a rope driven flexible three-finger hand for the selected manipulator and designs an independent control method for the manipulator. The complete hardware platform is designed for the subsequent mechanical arm base. In the study of visual servo control, it provides a powerful guarantee. (2) the control system of the manipulator consists of the manipulator and the visual system. Is a relatively complex control system, so the establishment of its software platform is very important. Visual servo systems need good visual sensors, After comparison, we choose the depth sensor Kinect and use computer vision library to process the images collected by the sensor, and then obtain the position information of the object to be grabbed in the workspace of the manipulator. For the control of robot arm, we operate and plan the manipulator under the framework of open source robot operating system, and add the control package to the robot arm by using the robot motion planning module. At the same time, in order to satisfy the intelligent control of the manipulator, the kinematics analysis of the manipulator is carried out in this paper. It includes forward kinematics and inverse kinematics and obtains its Jacobian matrix. Based on the derivation of dynamic equations of manipulator, the RBF neural network oriented to workspace is introduced into the control of manipulator. The Simulink simulation module of Matlab is used to simulate the application of the algorithm in the manipulator and the simulation results of the RBF control algorithm are analyzed. Finally, the specific working mode of the visual feedback system is introduced. The model of manipulator is transformed by URDF so that it can display its state through Rviz so as to help us detect the state of manipulator when grasping. Finally, we use the manipulator system built in this paper to carry out grasping, precision measurement test and line drawing test. It is proved that the manipulator control system based on visual feedback can accurately and stably grab the object and the RBF neural network control algorithm can improve the control accuracy and stability of the manipulator. The research system of the manipulator is established in the aspects of software to control algorithm operation platform, which lays a solid foundation for the further research and intelligence of the manipulator. This paper has 79 charts, 7 tables and 102 references.
【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP241

【参考文献】