Accuracy improvement with checkerboards in SLAM-integrated kinematic calibration (SKCLAM) for industrial manipulators

Abstract

Today, offline programing is available on industrial manipulators and used in many factories. However, the absolute accuracy of robots is insufficient due to manufacturing and assembly tolerances, thus it is necessary to perform kinematic calibration in advance. Most of existing calibration methods require a large amount of measurements just to calibrate a robot. In addition, when programing the robot offline, it is also important to perform motion planning for safe and efficient operations. Appropriate motion planning enables the robot to move along the optimal route taking obstacle avoidance and cycle time reduction into account. For this purpose, the accurate measurement of the environment around the robot is required, but existing methods require a lot of work and time. Therefore, in order to reduce the time and effort required for kinematic calibration and environmental mapping, we developed a SKCLAM (Simultaneous Kinematic Calibration, Localization And Mapping) method, in which kinematic calibration and environmental mapping are performed simultaneously for a manipulator with an RGB-D sensor attached to its hand. In this study, in order to improve the accuracy of the SKCLAM method, we introduced checkerboard patterns. We verified the SKCLAM method with checkerboard patterns in both a virtual environment and a real environment. The results showed that the effectiveness and some limitations of our current implementation.