Adaptive feedforward control of a collaborative industrial robot manipulator using a novel extension of the Generalized Maxwell-Slip friction model

Abstract

Collaborative industrial robots often use strain-wave transmissions which display a highly nonlinear behavior. In particular, the friction torque depend on the load torque and the hysteresis characteristics were recently found to depend on the joint angular position.This paper presents a novel extension of the Generalized Maxwell-Slip friction model to describe said phenomena in a combined framework. The method overcomes the discontinuity around zero velocity of existing models. Experiments on the Universal Robots UR5e manipulator show superior performance in terms of torque prediction accuracy and tracking performance of the proposed method.An adaptive feedforward friction compensator is proposed based on the extended Generalized Maxwell-Slip friction model to compensate the time-variations of the Coulomb and viscous friction due to, respectively, wear and mispredictions of the lubricant temperature. The adaptive estimator relies on the sensing hardware readily available in the joints of the Universal Robots manipulators, i.e. two absolute rotary encoders; one at each side of the transmission, current sensing for the electric actuator, and a temperature sensor. Results show a considerable reduction of the torque prediction error and tracking error.