Contrastive Imitation Learning-Based Scheduling Toward Deterministic Coordination of Parallel Flows in TSN-Enabled IIoT
IEEE Transactions on Networking · 2026 · 34 · 4716-4731
DOI: 10.1109/ton.2026.3682855 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Networking, 2026. Verfasst von Zhenrui Cao und 4 weiteren Autor:innen. Die Arbeit nennt 47 Literaturangaben. Gefördert von: National Science Fund for Distinguished Young Scholars of China, National Natural Science Foundation of China, Major Science and Technology Projects in Qinghai Province.
Abstract (Originalsprache)
In Time-Sensitive Networking (TSN)-enabled Industrial Internet of Things (IIoT), data for industrial applications is generated through task executions and delivered via flow transmissions, with task-flow scheduling offering computational and bandwidth resources. Distributed applications emphasize data acquisition via parallel flows, requiring deterministic coordination of parallel flows, i.e., strict jitter constraints on their delivery completions. However, existing studies lack deterministic guarantees for such coordination, leading to unreliable data acquisition for applications. To this end, we propose CA-TFS, a parallel-flow Coordination-Aware framework for Task-Flow Scheduling based on Contrastive Imitation Learning (CIL). We first develop a task-flow scheduling model that accommodates the parallel-flow coordination to meet the data acquisition needs for a wide range of applications. Next, we devise CIL to train an agent to identify high-quality scheduling solutions in the complex solution space, by fostering the agent to learn optimal solution trajectories revealed by Branch-and-Bound. Moreover, we tailor an incremental learning method for CIL, enabling the agent to generalize to large-scale task-flow scheduling. Extensive simulations validate CA-TFS’s superiority in ensuring deterministic parallel-flow coordination and meeting applications’ data acquisition needs, with its practicality confirmed via testbed evaluation in TSN-enabled IIoT.
Förderung
- National Science Fund for Distinguished Young Scholars of China
- National Natural Science Foundation of China
- Major Science and Technology Projects in Qinghai Province
- Fundamental Research Funds for the Central Universities
- Natural Science Foundation of Tianjin
Zitieren
APA
Cao, Z., Qiu, T., Zhou, X., Huang, M., & Wang, X. (2026). Contrastive Imitation Learning-Based Scheduling Toward Deterministic Coordination of Parallel Flows in TSN-Enabled IIoT. IEEE Transactions on Networking, 34, 4716-4731. https://doi.org/10.1109/ton.2026.3682855
Vancouver
Cao Z, Qiu T, Zhou X, Huang M, Wang X. Contrastive Imitation Learning-Based Scheduling Toward Deterministic Coordination of Parallel Flows in TSN-Enabled IIoT. IEEE Transactions on Networking. 2026;34:4716-4731. doi:10.1109/ton.2026.3682855
BibTeX
@article{Cao2026,
author = {Cao, Zhenrui and Qiu, Tie and Zhou, Xiaobo and Huang, Min and Wang, Xingwei},
title = {Contrastive Imitation Learning-Based Scheduling Toward Deterministic Coordination of Parallel Flows in TSN-Enabled IIoT},
journal = {IEEE Transactions on Networking},
year = {2026},
volume = {34},
pages = {4716--4731},
doi = {10.1109/ton.2026.3682855}
}
EUVIMED