A calibration procedure for a hand eye system

In the next generation model based robot systems, absolute positioning accuracy of a robot is becoming more important to effectively utilize advanced offline teaching and online programming techniques. This paper describes a calibration procedure for a hand-eye system with a multijoint direct-drive manipulator and a 3-D visual sensor system. A new technique to find optimal calibration parameters of the manipulator by using a singular value decomposition of a Jacobian matrix is proposed. The 3-D visual sensor system consisting of a laser pointer and a TV camera is also calibrated by estimating unknown model's parameters based on Newton method. Calibration of a hand-eye system is accomplished by determining external parameters of the manipulator relative to the world coordinate frame of the 3-D visual sensor system. As the result of calibration experiments, average positioning error of the hand-eye system is about 0.72 mm.<
>