EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Competition-Aware Deep Reinforcement Learning Framework for Practical Flexible Job Shop Scheduling

Yanqing Zhao, Yongze Ma, Chuanchen Wang, Yi Hu, Sifang Feng

Applied Sciences · 26.05.2026 · 16 (11) · 5340

DOI: 10.3390/app16115340 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Applied Sciences, online erschienen am 26.05.2026. Verfasst von Yanqing Zhao und 4 weiteren Autor:innen. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Ministry of Industry and Information Technology.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The flexible job shop scheduling problem (FJSP) is a typical combinatorial optimization problem in smart manufacturing. Although existing methods have considered machine competition relationships, they lack explicit structured modeling of machine competition relationships induced by candidate operations and are not systematically integrated across state representation, representation learning, and decision-making processes. To address this, this paper proposes a competition-aware dual-attention deep reinforcement learning method. We construct a dynamic heterogeneous graph representation, where machine competition is modeled as state-dependent edges instantiated via a 3D competition tensor, transforming machine competition relationships into structured information, thereby enhancing the model’s ability to characterize complex resource competition patterns. On this basis, we have designed the Competition-Aware Dual-Attention Network (CADAN), which injects competition information into both the attention computation and representation learning processes via a dual-path mechanism, enabling more expressive modeling of machine competition relationships, and which introduces a head-wise competition bias to capture heterogeneous competition patterns. Furthermore, we have developed an adaptive decision head to refine the scores of candidate actions. Our experimental results demonstrate that the proposed method outperforms classical dispatching rules and achieves competitive or superior performance compared with representative evolutionary and learning-based methods on synthetic datasets, public benchmark datasets, and a real-world industrial machining scenario involving mechanical transmission components.

Förderung

  • Ministry of Industry and Information Technology

Zitieren

APA

Zhao, Y., Ma, Y., Wang, C., Hu, Y., & Feng, S. (2026). A Competition-Aware Deep Reinforcement Learning Framework for Practical Flexible Job Shop Scheduling. Applied Sciences, 16(11), 5340. https://doi.org/10.3390/app16115340

Vancouver

Zhao Y, Ma Y, Wang C, Hu Y, Feng S. A Competition-Aware Deep Reinforcement Learning Framework for Practical Flexible Job Shop Scheduling. Applied Sciences. 2026;16(11):5340. doi:10.3390/app16115340

BibTeX
@article{Zhao2026,
  author = {Zhao, Yanqing and Ma, Yongze and Wang, Chuanchen and Hu, Yi and Feng, Sifang},
  title = {A Competition-Aware Deep Reinforcement Learning Framework for Practical Flexible Job Shop Scheduling},
  journal = {Applied Sciences},
  year = {2026},
  volume = {16},
  number = {11},
  pages = {5340},
  doi = {10.3390/app16115340}
}
Kostenlos starten →