EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

MLP-Based Inverse Kinematics for Efficient Pick and Place Motion Generation of 6-DOF Robotic Manipulators

Andara Dwi Hermawan, Aufaclav Zatu Kusuma Frisky, Andi Dharmawan

International Journal of Robotics and Control Systems · 30.03.2026 · 6 (2) · 980-998

DOI: 10.31763/ijrcs.v6i2.2463 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Robotics and Control Systems, online erschienen am 30.03.2026. Verfasst von Andara Dwi Hermawan und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Control systems for six degree of freedom (6-DOF) robotic manipulators in pick and place tasks commonly rely on classical inverse kinematics (IK) solvers that are iterative, sensitive to initialization, and can become unreliable near workspace boundaries. These limitations may reduce motion stability in repetitive execution and increase computation overhead due to repeated numerical refinement. This paper proposes a Multi-Layer Perceptron (MLP)-based inverse kinematics approach that provides deterministic feed-forward prediction while enforcing feasibility through joint-limit constraints. Although the robot is described as a 6-DOF manipulator, the IK mapping in this study targets five actuated revolute joints, while the sixth joint corresponds to the gripper and is excluded from the learning target. An end-to-end pipeline is developed covering URDF-based dataset generation, constrained learning with a bounded output layer, and deployment in both simulation and physical experiments. The proposed model is evaluated using end-effector point testing and pick and place motion execution in PyBullet, compared with a Jacobian-based numerical IK baseline implemented in IKPy (warm-start), the proposed method improves the success rate from 40% to 65% (?=10mm), reduces the mean Cartesian position error from 18.38 mm to 4.28 mm, and decreases pick/place end points error from 44.2/53.86 mm to 13.8/25.6 mm, respectively. The average runtime per target is reduced from 20 ms to 7.46 ms. These results indicate that the proposed MLP-based IK offers a practical accuracy efficiency trade off for real time pick and place applications.

Zitieren

APA

Hermawan, A. D., Frisky, A. Z. K., & Dharmawan, A. (2026). MLP-Based Inverse Kinematics for Efficient Pick and Place Motion Generation of 6-DOF Robotic Manipulators. International Journal of Robotics and Control Systems, 6(2), 980-998. https://doi.org/10.31763/ijrcs.v6i2.2463

Vancouver

Hermawan AD, Frisky AZK, Dharmawan A. MLP-Based Inverse Kinematics for Efficient Pick and Place Motion Generation of 6-DOF Robotic Manipulators. International Journal of Robotics and Control Systems. 2026;6(2):980-998. doi:10.31763/ijrcs.v6i2.2463

BibTeX
@article{Hermawan2026,
  author = {Hermawan, Andara Dwi and Frisky, Aufaclav Zatu Kusuma and Dharmawan, Andi},
  title = {MLP-Based Inverse Kinematics for Efficient Pick and Place Motion Generation of 6-DOF Robotic Manipulators},
  journal = {International Journal of Robotics and Control Systems},
  year = {2026},
  volume = {6},
  number = {2},
  pages = {980--998},
  doi = {10.31763/ijrcs.v6i2.2463}
}
Kostenlos starten →