基于图像的无标定混合双目机械手视觉伺服

发布时间：2018-05-05 14:24

本文选题：图像雅可比 + 混合双目　；参考：《重庆大学》2016年硕士论文

【摘要】：机器人视觉伺服是机器人研究领域的一个重要部分,其主要核心是对视觉系统的反馈信息进行处理,以此驱动机器人运动,完成机器人的定位、跟踪等控制任务。机器人是一个高度非线性的系统,并且视觉系统内部也存在非线性的映射关系。因此,机器人视觉伺服控制的一个的难点在于建立机器人和视觉系统之间的映射关系。本文以图像雅可比矩阵描述上述两者间的映射关系,采用了“眼固定”和“眼在手”相结合的布局方式,分别以点和角度为图像特征,组成图像雅可比矩阵。采用最小二乘法估计图像雅可比矩阵的初值,作为在线更新的基础,以卡尔曼滤波递推估计算法来进行图像雅可比矩阵的在线辨识更新,主要工作如下:(1)提出了基于图像的无标定混合双目机械手六自由度视觉伺服控制方法。对于机械手在任务空间中的位姿定位任务,采用基于图像的视觉伺服闭环控制方法,用卡尔曼滤波方法在线估计图像雅可比,通过图像雅可比提供的手眼映射关系,在图像平面设计期望的图像特征位置,通过图像特征偏差驱动、引导机械手在任务空间实现位姿定位的任务。设计了PID机械手视觉伺服控制器,并对控制器的稳定性条件进行了分析。(2)采用澳大利亚科学家Peter Corke开发的免费开源的机器人工具箱(9.9版本),结合机器视觉工具箱(3.3版本),在Matlab2014a环境下搭建了视觉伺服仿真平台。以“眼在手”和“眼固定”相结合的形式布局摄像头,并设计了基于图像的无标定机械手六自由度视觉伺服仿真对比试验。(3)利用参与搭建的由DENSO VS 6556GM六自由度工业机械手,RC7M控制器和大恒水星系列工业相机等设备组成的平台,以点和角度作为图像特征,用常值图像雅可比结合关节角限制,进行了实物试验。仿真试验结果表明,本文提出的基于图像的无标定混合双目机械手六自由度视觉伺服控制方法,能够有效、准确的完成机械手六自由度的位姿定位任务,验证了方法的可行性和有效性。同时,实物实验采用的方法也较好的实现了基于图像的无标定混合双目机械手视觉伺服位姿定位任务。
[Abstract]:Robot vision servo is an important part of robot research field. Its main core is to process the feedback information of vision system, so as to drive the robot motion, complete the robot positioning, tracking and other control tasks. Robot is a highly nonlinear system, and the vision system also has nonlinear mapping relationship. Therefore, one of the difficulties of robot visual servo control is to establish the mapping relationship between robot and vision system. In this paper, the image Jacobian matrix is used to describe the mapping relationship between the above two. The image Jacobian matrix is composed of "fixed eye" and "eye in hand" layout. The points and angles are taken as the image features, respectively, to form the image Jacobian matrix. The least square method is used to estimate the initial value of image Jacobian matrix, and the Kalman filter recursive estimation algorithm is used to estimate the initial value of image Jacobian matrix. The main work is as follows: 1) an image based visual servo control method for uncalibrated hybrid binocular manipulator with six degrees of freedom is proposed. For the position and pose orientation task of manipulator in the task space, the image based visual servo closed-loop control method and Kalman filter method are used to estimate the image Jacobian online, and the hand-eye mapping relationship provided by the image Jacobian is used. The desired image feature position is designed in the image plane and driven by the image feature deviation to guide the manipulator to realize the position orientation task in the task space. The visual servo controller of PID manipulator is designed. The stability condition of the controller is analyzed. (2) A visual servo simulation platform under Matlab2014a environment is built by using the free and open source robot toolbox developed by Australian scientist Peter Corke and the machine vision toolbox with 3.3 version. In the form of a combination of "eye in hand" and "eye fixation", The visual servo simulation contrast test of uncalibrated manipulator based on image is designed. The platform is composed of DENSO vs 6556GM 6-DOF industrial manipulator RC7M controller and Daheng Mercury series industrial camera. Point and angle are taken as image features, and a real object experiment is carried out by using constant image Jacobian combined with joint angle restriction. The simulation results show that the six degree of freedom visual servo control method of uncalibrated hybrid binocular manipulator proposed in this paper can effectively and accurately complete the position and orientation task of the manipulator with six degrees of freedom. The feasibility and effectiveness of the method are verified. At the same time, the method used in the physical experiment also achieves the task of uncalibrated hybrid binocular manipulator visual servo position and pose location based on image.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TP241

【相似文献】