Real-time observer designs for elastic-joint industrial robots: Experimental comparison and new strategies

Abstract

Due to unavoidable compliance in the drive trains of the individual links, all industrial robots exhibit some elasticity in their respective joints. This elasticity causes a discrepancy between the measured motor position and the actual link position of the robot. For highly dynamic movements of the robot, or in the case of heavy payloads, this can lead to significant deviations between the desired and actual end-effector pose of the robot. One approach to mitigate the resulting task performance degradation is using suitable state estimation strategies for elastic-joint robots. While critical for task performance, the link positions are typically not separately measured for industrial robots. Instead, standard robot control systems rely on the measured motor positions. To this end, this work studies the state estimation problem of elastic-joint industrial robots and compares the performance of different observers. The comparison is performed on an industrial robot with six degrees of freedom (6-DOF) using a laser tracker measurement of the end-effector movement as a ground truth. Four observer concepts are state-of-the-art strategies from the literature, which are briefly summarized first. Subsequently, two novel estimation strategies are proposed. The first exploits encoder and angular velocity measurements, and the second is an estimator that combines encoder, acceleration, and angular velocity measurements. To the best of the authors’ knowledge, this is the first work that compares the performance of real-time observers for elastic-joints industrial robots with 6-DOF in an experimental setup. The properties of all observers are discussed in terms of estimation accuracy, sensor configuration, computational complexity, and required knowledge of model parameters. The experiments show that the novel estimation strategies improve the position estimation accuracy of the end-effector by up to 72 %.