EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Zero-Shot Payload Adaptation for Robotic Manipulators via Domain-Randomized Deep Reinforcement Learning

Saifullah Samo, Sameh EZZT Mohamed Ali Bdran, Raheel Ahmed Nizamani

Mehran University Research Journal of Engineering and Technology · 01.04.2026 · 45 (2) · 122-133

DOI: 10.22581/0621 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Mehran University Research Journal of Engineering and Technology, online erschienen am 01.04.2026. Verfasst von Saifullah Samo und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Reliable robotic manipulator control under unforeseen payload changes has been studied because such variation creates a major challenge for consistent automated operation. Traditional model-based approaches have been viewed as unreliable since parameter uncertainty weakens system response, while adaptive methods have been reported with slow convergence behaviour. A key gap has been described as the lack of one stable non adaptive policy that ensures steady reliability across many physical parameters. This study has been designed to create and verify a robust control policy that supports strong zero-shot generalization for unseen payload conditions. The method has been formed by linking deep reinforcement learning with domain randomization inside a detailed simulator that trains an agent through Proximal Policy Optimization. The central innovation has been presented as the ability to produce generalized control across many randomized dynamic models used during simulated training cycles. Simulation results have been reported with superior trajectory tracking and stable system motion across unseen payloads, showing reduced error values in varied tests. Measured outcomes have shown low mean error of 0.032 rad with only a 20% rise from baseline and steady energy use.

Zitieren

APA

Samo, S., Bdran, S. E. M. A., & Nizamani, R. A. (2026). Zero-Shot Payload Adaptation for Robotic Manipulators via Domain-Randomized Deep Reinforcement Learning. Mehran University Research Journal of Engineering and Technology, 45(2), 122-133. https://doi.org/10.22581/0621

Vancouver

Samo S, Bdran SEMA, Nizamani RA. Zero-Shot Payload Adaptation for Robotic Manipulators via Domain-Randomized Deep Reinforcement Learning. Mehran University Research Journal of Engineering and Technology. 2026;45(2):122-133. doi:10.22581/0621

BibTeX
@article{Samo2026,
  author = {Samo, Saifullah and Bdran, Sameh EZZT Mohamed Ali and Nizamani, Raheel Ahmed},
  title = {Zero-Shot Payload Adaptation for Robotic Manipulators via Domain-Randomized Deep Reinforcement Learning},
  journal = {Mehran University Research Journal of Engineering and Technology},
  year = {2026},
  volume = {45},
  number = {2},
  pages = {122--133},
  doi = {10.22581/0621}
}
Kostenlos starten →