Pose Error Estimation Using a Cylinder in Scanner-Based Robotic Belt Grinding

Abstract

Scanner-based robotic grinding has shown great potential for replacing the manual method to achieve efficient and automatic manufacturing. However, its application has been limited by the grinding quality, which is affected by the pose errors between the robot and other workcells (a scanner, a workpiece, and a tool). To improve the pose accuracy, this article proposes a novel estimation method for above three pose errors, where a cylinder is used as the calibration object. Using differential kinematics, the relationship between the hand-eye pose error and the reconstructed error is built. The hand-eye pose error is estimated by scanning and reconstructing a cylinder. An iterative method is presented to find the optimized orientation error vector, avoiding the uncertain error of multiple solutions using the Schmidt method. Based on the speed adjoint transformation, a mathematical model between the workpiece/tool pose errors and grinding error is built. Then, an experimental estimation approach for the workpiece/tool pose errors is presented by shape matching the measured points of the grinded cylinder with the design model. Unlike the traditional static construction method, this method uses large-scale measured points with high measurement accuracy to improve the estimation accuracy and stability and can estimate the pose error caused during the dynamic grinding process, such as vibration and force deformation. Finally, pose error estimation and compensation experiments are performed to verify the feasibility of the proposed method.