基于Adam优化算法的双目机器人手眼标定方法

Adam-based Hand-eye Calibration Method for Robot with Binocular Vision

在智能制造领域,视觉机器人应用前景十分广阔。视觉机器人的手眼标定精度直接关系到机器人的后续作业精度。为了进一步提高机器人的手眼标定精度,现提出一种基于Adam优化算法的双目Eye-to-Hand型机器人的手眼标定方法。根据多体运动学理论,建立了6DOF机器人手眼标定数学模型,以Halcon输出的手眼标定矩阵为初始值,采用Adam优化算法对目标函数进行迭代求解,将由优化前后手眼矩阵得到的两组机器人末端坐标系的位姿分别与从示教器得到的位姿作差值,并取Frobenius范数。结果表明:相机标定误差为0.089个像素,优化后的Frobenius范数平均值小于优化前,且一致性好。

The application prospects of robots with vision navigation are very broad in intelligent manufacturing field,the accuracy of hand-eye calibration of vision robots is directly related to the subsequent operations accuracy of the robot.In order to further improve the accuracy of the hand-eye calibration of the robot,a binocular Eye-to-Hand robot hand-eye calibration method was proposed based on Adam optimization algorithm.According to the theory of multi-body kinematics,the hand-eye calibration model of a 6 DOF robot was established.Taking the hand-eye calibration matrix obtained by Halcon as the initial input,the objective function was solved iteratively by using Adam optimization algorithm,and the subtraction operation was run between two sets of terminal postures obtained before and after optimization and the posture of the terminal obtained by teaching box,then their Frobenius norms were calculated.The experimental results show that the error of camera calibration is 0.089 pixels,the average of Frobenius norm after optimization is smaller than that before optimization,and its consistency is quite well.