Abstract: Considering the problem that the conventional robot self-calibration device based on geometric constraints can only measure and calibrate partial robot poses in the local robot workspace, a novel portable robot self-calibration device is proposed, which consists of a ball center measuring device installed at the flange of robot and a movable ball bar. By utilizing spherical and distance constraints, it can measure and calibrate a large range of robot targets in large workspace, which improves the reliability of the calibration results. According to the single-ball and double-ball arrangements of the movable ball bar, two robot self-calibration models as well as algorithms based on vector difference and distance difference are established respectively. By using the local product of exponential (POE) formula and introducing position adjoint transformation matrix, both the self-calibration models are simplified, which reduces the linearizing calculation of the kinematic equations. Finally, a simulation experiment on a 6-DOF (degree of freedom) serial manipulator is implemented and the simulation results show that both the proposed self-calibration algorithms can converge quickly, which verifies that both the calibration algorithms are effective and robust.