A Kinematic Error Controller for Real-Time Kinematic Error Correction of Industrial Robots

Abstract

Industrial robots are being used more and more for manufacturing applications that require accuracy beyond what can be obtained from joint measurement. While offline calibration techniques such as volumetric error compensation can be used to correct robot kinematic error, these methods are unable to compensate for robot deformations caused by changing tool loads during the manufacturing operation. This paper explores the use of a real-time robot kinematic error compensation technique where an external high-precision feedback sensor (in this case a laser tracker) directly measures the robot kinematic error and corrections are implemented during processing. A robot kinematic error model is constructed to describe the difference between the programmed trajectory of the robot’s last link and the actual trajectory based on the laser tracker measurement of the 6DoF sensor attached to the last link of the robot. The compensation scheme developed in this paper requires the synchronization of the robot encoder and laser tracker sensor measurements, which is accomplished in the Robot Operating System (ROS). The previously developed kinematic error observer is briefly discussed and a kinematic error compensation that adjusts the robot’s reference path in real time is created in this paper. A discussion of the system (i.e., Yaskawa/Motoman MH180 industrial robot and Automated Precision Inc. laser tracker) hardware components and the software architecture utilized in the experiments conducted in this paper are provided. Initial experimental studies are conducted to determine delays in the feedback measurement, which was 30 ms, explore the effects of controller gain on the system performance, and characterize the noise in the feedback measurement. The controller had an overdamped response with a settling time of 8.758 s. Subsequent experiments investigated the effect of robot velocity on tracking and the ability of the controller to reject a constant force disturbance. It was found that the kinematic error increased linearly as the robot velocity increased, and the controller was able to reject a constant force disturbance of 45 lb within the designed settling time.