by Nadia Schillreff, Frank Ortmeier
Abstract:
A learning-based robot kinematic calibration approach based on the product-of-exponentials (POE) formula and Adjoint error model is introduced. To ensure high accuracy this approach combines the geometrical and non-geometrical influences like for e.g. elastic deformations without explicitly defining all physical processes that contribute to them using a polynomial regression method. By using the POE formula for kinematic modeling of the manipulator it is ensured that kinematic parameters vary smoothly and used method is robust and singularity-free. The introduced error parameters are presented in the form of Adjoint transformations on nominal joint twists. The calibration process then becomes finding a set of polynomial functions using regression methods that are able to reflect the actual kinematics of the robot. The proposed method is evaluated on a dataset obtained using a 7-DOF manipulator (KUKA LBR iiwa 7 R800). The experimental results show that this approach significantly reduc es positional errors of the robotic manipulator after calibration.
Reference:
Learning-based Kinematic Calibration using Adjoint Error Model (Nadia Schillreff, Frank Ortmeier), In Proceedings of the 15th International Conference on Informatics in Control, Automation and Robotics - Volume 2: ICINCO (SciTePress, ed.), 2018.
Bibtex Entry:
@inproceedings{schillreff_learning-based_2018,
	address = {Porto, Portugal},
	title = {Learning-based {Kinematic} {Calibration} using {Adjoint} {Error} {Model}},
	isbn = {978-989-758-321-6},
	doi = {10.5220/0006870403820389},
	abstract = {A learning-based robot kinematic calibration approach based on the product-of-exponentials (POE) formula and Adjoint error model is introduced. To ensure high accuracy this approach combines the geometrical and non-geometrical influences like for e.g. elastic deformations without explicitly defining all physical processes that contribute to them using a polynomial regression method. By using the POE formula for kinematic modeling of the manipulator it is ensured that kinematic parameters vary smoothly and used method is robust and singularity-free. The introduced error parameters are presented in the form of Adjoint transformations on nominal joint twists. The calibration process then becomes finding a set of polynomial functions using regression methods that are able to reflect the actual kinematics of the robot. The proposed method is evaluated on a dataset obtained using a 7-DOF manipulator (KUKA LBR iiwa 7 R800). The experimental results show that this approach significantly reduc es positional errors of the robotic manipulator after calibration.},
	booktitle = {Proceedings of the 15th {International} {Conference} on {Informatics} in {Control}, {Automation} and {Robotics} - {Volume} 2: {ICINCO}},
	author = {Schillreff, Nadia and Ortmeier, Frank},
	editor = {{SciTePress}},
	month = aug,
	year = {2018},
	keywords = {modeling, Parameter Identification},
	pages = {382--389}
}