EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

EdgeOpt-Sched-CS: Cold-Start-Aware Dynamic Scheduling for Efficient DNN Inference at the Edge

Yuchang Gu iD, Diming Zhang iD, Taiyu Lu iD

Sensors · 15.05.2026 · 26 (10) · 3130

DOI: 10.3390/s26103130 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 15.05.2026. Verfasst von Yuchang Gu und 2 weiteren Autor:innen. Beteiligte Einrichtung: Ocean College, Jiangsu University of Science and Technology, Zhenjiang 212000, China. Die Arbeit nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Dynamic scheduling can improve the latency and memory efficiency of deep neural network inference on edge devices, but it often introduces cold-start overhead when a newly deployed model requires online profiling and policy adaptation before reaching stable performance. This paper proposes EdgeOpt-Sched-CS, a cold-start-aware extension of dynamic graph scheduling for edge inference. The key idea is to initialize the scheduler of a target computation graph using scheduling knowledge transferred from structurally similar source graphs, instead of starting from a generic policy. EdgeOpt-Sched-CS constructs compact graph signatures, retrieves relevant source schedulers, and performs lightweight cold-start-aware online adaptation during early deployment. We evaluate the framework across representative device–model scenarios involving lightweight convolutional neural networks, transformer models, and quantized language-model workloads. The results show that EdgeOpt-Sched-CS reduces cumulative cold-start latency by 10.6–20.4% and shortens time-to-stability by 5.2–21.7%, while preserving the steady-state latency–memory behavior of the original dynamic scheduler with only small additional scheduling overhead. These findings indicate that scheduler initialization is an important optimization dimension for adaptive edge inference and that prior scheduling knowledge can be effectively reused across related computation graphs.

Schlagwörter

Stichwörter der Autor:innen

cold-start optimizationdynamic schedulingedge inferencegraph neural networksoperator schedulingtransfer learningwarm-start scheduling

Autor:innen und Institutionen

  • Ocean College, Jiangsu University of Science and Technology, Zhenjiang 212000, China.
  • School of Computer, Jiangsu University of Science and Technology, Zhenjiang 212000, China.

ORCID-Profile: Yuchang Gu, Diming Zhang, Taiyu Lu

Zitieren

APA

Gu, Y., Zhang, D., & Lu, T. (2026). EdgeOpt-Sched-CS: Cold-Start-Aware Dynamic Scheduling for Efficient DNN Inference at the Edge. Sensors, 26(10), 3130. https://doi.org/10.3390/s26103130

Vancouver

Gu Y, Zhang D, Lu T. EdgeOpt-Sched-CS: Cold-Start-Aware Dynamic Scheduling for Efficient DNN Inference at the Edge. Sensors. 2026;26(10):3130. doi:10.3390/s26103130

BibTeX
@article{Gu2026,
  author = {Gu, Yuchang and Zhang, Diming and Lu, Taiyu},
  title = {EdgeOpt-Sched-CS: Cold-Start-Aware Dynamic Scheduling for Efficient DNN Inference at the Edge},
  journal = {Sensors},
  year = {2026},
  volume = {26},
  number = {10},
  pages = {3130},
  doi = {10.3390/s26103130}
}
Kostenlos starten →