电子学报 ›› 2022, Vol. 50 ›› Issue (10): 2318-2328.DOI: 10.12263/DZXB.20201353

• 学术论文 • 上一篇    下一篇

可变环境下基于位姿变换矩阵的机器人无标定手眼协调方法

金紫凤, 潘思聪, 危辉()   

  1. 复旦大学计算机科学技术学院认知算法模型实验室,上海 200438
  • 收稿日期:2020-11-30 修回日期:2021-05-12 出版日期:2022-10-25
    • 通讯作者:
    • 危辉
    • 作者简介:
    • 金紫凤 女,1995年生,浙江嘉善人.现为复旦大学计算机科学技术学院硕士研究生.主要研究方向为智能机器人.E-mail: zfjin18@fudan.edu.cn
      潘思聪 男,1996年生,上海人.现为复旦大学计算机科学技术学院硕士研究生.主要研究方向为智能机器人.E-mail: 18210240033@fudan.edu.cn
      危 辉(通讯作者) 男,1971年生,江西南昌人.现为复旦大学计算机科学技术学院教授、博士生导师.主要研究方向为人工智能理论与技术、认知科学.
    • 基金资助:
    • 国家自然科学基金 (61771146)

Uncalibrated Hand Eye Coordination Method for Robot Based on Pose Transformation Matrix in Variable Environment

JIN Zi-feng, PAN Si-cong, WEI Hui()   

  1. Laboratory of Algorithms for Cognitive Models,School of Computer Science,Fudan University,Shanghai 200438,China
  • Received:2020-11-30 Revised:2021-05-12 Online:2022-10-25 Published:2022-10-11
    • Corresponding author:
    • WEI Hui

摘要:

智能机器人不同于工业流水线上固定任务的机器人,它们往往需要面对背景、目标物体形状、位置、姿态、尺寸的多种不确定性,这就要求机器人能够像人一样具有极好的手眼协调能力,能基于动态反馈临机调整自己的动作.本文采用基于RGB-D相机的三维深度信息采集手段,构建了一个用于无标定可变场景机器人手眼协调方法,它能够利用几何关系理解三维场景,实时跟踪并准确分割机械臂末端执行器和待抓取的物体,在没有3D模型的条件下计算末端执行器的姿态,通过计算夹持器的姿态以及夹持器和目标对象在相机坐标系中的相对位移向量之后,将该相对位移矢量从相机坐标系转换到夹持器坐标系,估计夹持器位姿变换矩阵,并将齐次变换矩阵输入到机械手逆运动学程序中,求解出机械臂对应的各关节扭角,从而驱动夹持器指向或移动到物体.这一方法硬件资源需求量少、计算速度快、实时性好.

关键词: 手眼协调, 无标定, 三维环境感知, 计算几何, 机器人

Abstract:

Intelligent robots are different from the fixed task robots on the industrial assembly line. They often need to face a variety of uncertainties, such as the background, the shape, position, posture and size of the target object. This requires that the robot can have excellent hand eye coordination ability like us, and adjust its actions based on dynamic feedback. In this paper, a 3D depth information acquisition method based on RGB-D camera is used to construct a hand eye coordination method for uncalibrated variable scene robot. It can understand the 3D scene by using geometric relations, track and segment the end effector and the object to be grasped in real time, calculate the pose of the end effector without 3D model, and calculate the pose of the gripper. After the relative displacement vector of the gripper and the target object in the camera coordinate system, the relative displacement vector is transformed from the camera coordinate system to the gripper coordinate system, and the pose transformation matrix of the gripper is estimated, and the homogeneous transformation matrix is input into the inverse kinematics program of the manipulator to solve the corresponding joint torsion angles of the manipulator, so as to drive the gripper to point or move to the object. This method has the advantages of less hardware resource requirement, fast computing speed and good real-time performance.

Key words: hand-eye coordination, uncalibrated, 3D environment perception, computational geometry, robot

中图分类号: