Multi-Agent Hierarchical Graph Attention Actor–Critic Reinforcement Learning
Entropy · 25.12.2024 · 27 (1) · 4
DOI: 10.3390/e27010004 ↗Kurz gefasst
Fachartikel in der Zeitschrift Entropy, online erschienen am 25.12.2024. Verfasst von Tongyue Li und 5 weiteren Autor:innen. Beteiligte Einrichtung: Academy of Military Sciences, Beijing 100097, China. Die Arbeit wurde bisher 17-mal zitiert (Crossref) und nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Multi-agent systems often face challenges such as elevated communication demands, intricate interactions, and difficulties in transferability. To address the issues of complex information interaction and model scalability, we propose an innovative hierarchical graph attention actor–critic reinforcement learning method. This method naturally models the interactions within a multi-agent system as a graph, employing hierarchical graph attention to capture the complex cooperative and competitive relationships among agents, thereby enhancing their adaptability to dynamic environments. Specifically, graph neural networks encode agent observations as single feature-embedding vectors, maintaining a constant dimensionality irrespective of the number of agents, which improves model scalability. Through the “inter-agent” and “inter-group” attention layers, the embedding vector of each agent is updated into an information-condensed and contextualized state representation, which extracts state-dependent relationships between agents and model interactions at both individual and group levels. We conducted experiments across several multi-agent tasks to assess our proposed method’s effectiveness, stability, and scalability. Furthermore, to enhance the applicability of our method in large-scale tasks, we tested and validated its performance within a curriculum learning training framework, thereby enhancing its transferability.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Academy of Military Sciences, Beijing 100097, China.
- Tianjin Artificial Intelligence Innovation Center (TAIIC), Tianjin 300450, China.
- College of Computer, National University of Defense Technology, Changsha 410073, China.
- School of Computer Science, Peking University, Beijing 100871, China.
ORCID-Profile: Tongyue Li, Dianxi Shi, Songchang Jin, Zhen Wang, Huanhuan Yang, Yang Chen
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Li, T., Shi, D., Jin, S., Wang, Z., Yang, H., & Chen, Y. (2024). Multi-Agent Hierarchical Graph Attention Actor–Critic Reinforcement Learning. Entropy, 27(1), 4. https://doi.org/10.3390/e27010004
Vancouver
Li T, Shi D, Jin S, Wang Z, Yang H, Chen Y. Multi-Agent Hierarchical Graph Attention Actor–Critic Reinforcement Learning. Entropy. 2024;27(1):4. doi:10.3390/e27010004
BibTeX
@article{Li2024,
author = {Li, Tongyue and Shi, Dianxi and Jin, Songchang and Wang, Zhen and Yang, Huanhuan and Chen, Yang},
title = {Multi-Agent Hierarchical Graph Attention Actor–Critic Reinforcement Learning},
journal = {Entropy},
year = {2024},
volume = {27},
number = {1},
pages = {4},
doi = {10.3390/e27010004}
}
EUVIMED