Vision based manipulator self-calibration with motion estimation

A new method of calibrating manipulator kinematic parameters based on a stereo-vision/manipulator system is presented. The calibration is done concurrently with the estimation of the motion of an object observed by stereo cameras mounted on the manipulator end-effector. The object may be fixed in space or moving in a known form of trajectory but with unknown parameters. Since the proposed method does not depend on special calibration tools, the method is referred to here as a “self-calibration” method. In this paper, the authors first establish a general theory on vision-based manipulator self-calibration. Then, a detailed self-calibration equation relating the noise and biases involved in a stereo-vision/manipulator system to an object pose estimation error is derived for real applications. Simulation results support the feasibility and effectiveness of the proposed self-calibration method