Off-Policy Reinforcement Learning for Efficient Knowledge Transfer in Cooperative Decision Frameworks
Global Media and Social Sciences Research Journal · 05.04.2026 · 7 (1) · 298-303
DOI: 10.71465/gmssrj176 ↗Kurz gefasst
Fachartikel in der Zeitschrift Global Media and Social Sciences Research Journal, online erschienen am 05.04.2026. Verfasst von Maximilian Schmidt und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Data inefficiency remains a bottleneck in training collaborative decision systems, particularly when task environments vary across domains. This study examines the use of off-policy reinforcement learning to enable efficient knowledge reuse across tasks. A soft actor-critic (SAC) framework is extended with a shared replay buffer and domain-adaptive importance sampling to support cross-task learning. The method is tested on a dataset of 10,700 decision trajectories spanning logistics optimization, task allocation, and sequential planning scenarios. Compared with on-policy baselines, the proposed approach reduces training sample requirements by 38.4% and improves average task performance by 17.2%. Transfer experiments further show that policies trained on one domain retain 81.5% of their performance when applied to unseen environments. The results demonstrate that off-policy learning provides a scalable solution for multi-domain collaborative optimization.
Zitieren
APA
Schmidt, M., Müller, J., & Fischer, L. (2026). Off-Policy Reinforcement Learning for Efficient Knowledge Transfer in Cooperative Decision Frameworks. Global Media and Social Sciences Research Journal, 7(1), 298-303. https://doi.org/10.71465/gmssrj176
Vancouver
Schmidt M, Müller J, Fischer L. Off-Policy Reinforcement Learning for Efficient Knowledge Transfer in Cooperative Decision Frameworks. Global Media and Social Sciences Research Journal. 2026;7(1):298-303. doi:10.71465/gmssrj176
BibTeX
@article{Schmidt2026,
author = {Schmidt, Maximilian and Müller, Johannes and Fischer, Leon},
title = {Off-Policy Reinforcement Learning for Efficient Knowledge Transfer in Cooperative Decision Frameworks},
journal = {Global Media and Social Sciences Research Journal},
year = {2026},
volume = {7},
number = {1},
pages = {298--303},
doi = {10.71465/gmssrj176}
}
EUVIMED