Modeling of Robotic Manipulators Using a Moment-Loading Kinematic Error Method

Abstract

Industrial robots are often are much less accurate than they are repeatable. There is significant research interest in using industrial robot arms for applications which require high accuracy such as robotic machining. A variety of robot calibration methods have been developed to improve robot accuracy including geometric methods, which identify best-fit offsets to Denavit-Hartenberg parameters (DH parameters) or equivalents. Other non-geometric calibration methods including Volumetric Error Compensation (VEC) use higher-order models that can capture complex kinematic errors arising from robot joint non-linearity. An extended VEC model incorporating deflection due to loading about the axis of each robot joint is introduced in this paper. Calibration performance was compared between a $3^{\mathrm{rd}}$ degree VEC model, a $6^{\mathrm{th}}$ degree VEC model, and the extended model incorporating robot deflection. The VEC model demonstrated a 99% gamma distribution position error of 1.85 mm, or approximately 4.9 times the Motoman MH180’s bi-directional position repeatability on a separate validation set of measurements. Using the extended model incorporating robot deflection improved the mean position error by 46% to 0.99 mm or to 2.6 times the bi-directional position repeatability of the Motoman MH180 robot.