Centralized Training with Decentralized Execution for Scalable Coordination in Distributed Decision Environments
Fundamental Scientific Reports in Multidisciplinary Areas · 05.04.2026 · 2 (01) · 87-93
DOI: 10.66238/fsrma58 ↗Kurz gefasst
Fachartikel in der Zeitschrift Fundamental Scientific Reports in Multidisciplinary Areas, online erschienen am 05.04.2026. Verfasst von Lukas Meier und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Scalability remains a limiting factor in collaborative decision systems as the number of agents increases. This study examines a centralized training with decentralized execution (CTDE) paradigm to improve coordination efficiency under large-scale settings. A QMIX-based value decomposition framework is employed, allowing joint action-value estimation during training while preserving independent execution policies. The approach is evaluated on 9,400 large-scale coordination tasks with agent populations ranging from 10 to 50. Compared with independent learning baselines, the proposed method increases global reward by 28.6% and improves coordination success rate from 68.5% to 82.9%. Notably, performance degradation remains below 5% as agent count scales, indicating strong robustness. These results demonstrate that value factorization under CTDE is effective for managing complexity in distributed decision systems
Zitieren
APA
Meier, L., Schmid, A., & Müller, D. (2026). Centralized Training with Decentralized Execution for Scalable Coordination in Distributed Decision Environments. Fundamental Scientific Reports in Multidisciplinary Areas, 2(01), 87-93. https://doi.org/10.66238/fsrma58
Vancouver
Meier L, Schmid A, Müller D. Centralized Training with Decentralized Execution for Scalable Coordination in Distributed Decision Environments. Fundamental Scientific Reports in Multidisciplinary Areas. 2026;2(01):87-93. doi:10.66238/fsrma58
BibTeX
@article{Meier2026,
author = {Meier, Lukas and Schmid, Anna and Müller, David},
title = {Centralized Training with Decentralized Execution for Scalable Coordination in Distributed Decision Environments},
journal = {Fundamental Scientific Reports in Multidisciplinary Areas},
year = {2026},
volume = {2},
number = {01},
pages = {87--93},
doi = {10.66238/fsrma58}
}
EUVIMED