EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

WARA: Toward Automated Wireless Optimization Research With Closed-Loop LLM Agents

Yuan Guo, Yilong Chen, Chao Hu, Xianghao Yu, Liang Hong, Jie Xu

IEEE Open Journal of the Communications Society · 2026 · 7 · 13509-13524

DOI: 10.1109/ojcoms.2026.3745043 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Open Journal of the Communications Society, 2026. Verfasst von Yuan Guo und 5 weiteren Autor:innen. Die Arbeit nennt 63 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Shenzhen Fundamental Research Program, Hong Kong Research Grants Council.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Large language model (LLM) agents are increasingly capable of external tool use, code generation and execution, intermediate artifact inspection, and iterative output revision, creating new opportunities for automating scientific and engineering research. To the best of our knowledge, this paper presents the first end-to-end autoresearch framework for the wireless domain, with a particular focus on wireless resource allocation optimization, an essential area for characterizing the fundamental performance limits of wireless systems and enhancing their practical performance under dynamic channel and network conditions. Specifically, we propose the Wireless AutoResearch Agent (WARA), a closed-loop multi-agent system for automated wireless optimization research. Given only an initial research topic as input, WARA decomposes the research workflow into three phases: 1) research gap identification and problem proposal, 2) wireless optimization problem modeling, solution algorithm design, and experimentation, and 3) construction of research deliverables. For each phase, we design artifact-mediated control mechanisms, where declared upstream artifacts are consumed as inputs to produce structured outputs for downstream use. A controller-managed gate is then applied in each phase to validate the generated artifacts and ensure consistency among models, algorithms, experiments, and claims. When an artifact fails validation, only the corresponding artifact needs to be repaired, rather than restarting the entire workflow. Furthermore, we present a representative wireless resource allocation case study to demonstrate how WARA converts an initial topic into a complete research package with executable evidence and a synthesized technical manuscript. We also design a structured LLM-based ScoringAgent to evaluate manuscript-level research validity and optimization research maturity, and the comparative evaluations show that WARA substantially outperforms one-shot LLM generation while approaching the quality profile of recently accepted peer-reviewed technical papers. These results demonstrate that closed-loop artifact control is a promising path toward end-to-end LLM-assisted wireless optimization research. The source code is available at https://github.com/guoyuan-dotcom/WARA_CUHKSZ

Förderung

  • National Natural Science Foundation of China
  • Shenzhen Fundamental Research Program
  • Hong Kong Research Grants Council

Zitieren

APA

Guo, Y., Chen, Y., Hu, C., Yu, X., Hong, L., & Xu, J. (2026). WARA: Toward Automated Wireless Optimization Research With Closed-Loop LLM Agents. IEEE Open Journal of the Communications Society, 7, 13509-13524. https://doi.org/10.1109/ojcoms.2026.3745043

Vancouver

Guo Y, Chen Y, Hu C, Yu X, Hong L, Xu J. WARA: Toward Automated Wireless Optimization Research With Closed-Loop LLM Agents. IEEE Open Journal of the Communications Society. 2026;7:13509-13524. doi:10.1109/ojcoms.2026.3745043

BibTeX
@article{Guo2026,
  author = {Guo, Yuan and Chen, Yilong and Hu, Chao and Yu, Xianghao and Hong, Liang and Xu, Jie},
  title = {WARA: Toward Automated Wireless Optimization Research With Closed-Loop LLM Agents},
  journal = {IEEE Open Journal of the Communications Society},
  year = {2026},
  volume = {7},
  pages = {13509--13524},
  doi = {10.1109/ojcoms.2026.3745043}
}
Kostenlos starten →