A Framework of <i>In Situ</i> Model Error Compensation for Adaptive Robotic Task Execution

Abstract

In precision robotic applications, inaccuracy in workpiece geometry has been a common problem to the precise processing of the workpiece. Due to manufacturing defects and workpiece deformation, the actual geometry of the workpiece deviates from its nominal 3D CAD model which is defined as model error. For many of the existing industrial robotic applications today, the robot path for processing the workpiece is planned based on the nominal 3D CAD model of the workpiece. Hence, the model error of the workpiece leads to error in the robot path planning eventually inducing inaccurate processing. To enhance the accuracy of the robot in processing the workpiece, a framework for in-situ model error compensation has been proposed. Prior to the processing of each workpiece, the proposed technique employs 3D optical laser scanning technology to capture the actual 3D model of the workpiece and compares it with the nominal model to establish the model errors. The nominal path of the robot initially created based on the nominal CAD model is then modified according to the model error. Thus, this step performs the in-situ model error compensation making the robotic task adaptive to the actual workpiece geometry. Experiments have been conducted to verify the proposed framework and an accuracy of up to 30 micrometers has been achieved in model error identification and compensation. The proposed technique can be employed in applications such as high precision robotic tasks, where accuracy of task execution is an important factor.