Target ball localization for industrial robots based on binocular stereo vision

Purpose The purpose of this paper is to present a high-precision method for localizing the Target Ball of an industrial robot, using Binocular Stereo Vision (BSV). Design/methodology/approach The steps for modeling the kinematics of a flexible robot include three steps. First, the kinematic model was developed, and robot motion tests were conducted, which is multiple sets of static joint angle motions. The kinematic parameters were calibrated based on a laser tracker and BSV, respectively. Second, the dynamics model was developed and operate the robot to perform motions, which is Fourier series excitation trajectories. The dynamics parameters were calibrated. Third, the static joint angles were brought into the dynamics model to calculate the external forces and moments of each joint. The stiffness parameters were calibrated. Findings The experimental results show that after calibrating the kinematic parameters of the Target Ball, the Target Ball positioning error approximately 4.597 mm. After calibrating the kinematic parameters of the Spherically Mounted Retro and stiffness parameters, the Target Ball positioning error is reduced from about 6.962 mm to approximately 4.956 mm, which is close to 4.597 mm. Research limitations/implications The limitation of the approach is that, kinematic model and dynamics model are both required. Therefore, the high openness degree of robot is required, the robot needs to support path planning functions and joint moment extraction. However, some mature commercial robots do not support this. Originality/value In this paper, the stiffness model could effectively replace the calibration of the kinematic parameters of the load.