Human-like Communication Strategies for Improved Multi-Agent Reinforcement Learning
Keywords: RL, MARL, communication, RLHF
Abstract: Multi-Agent Reinforcement Learning (MARL) has seen significant progress in recent years, enabling multiple agents to coordinate and optimize their actions in complex environments. However, integrating effective communication protocols into MARL frameworks remains a challenge, as it introduces issues such as increased state space dimensionality, lack of stationarity, and the need for interpretability. Inspired by human communication, which relies on prior knowledge, contextual awareness, and efficient information exchange, we propose a novel framework for incorporating human-like communication strategies to enhance the learning process. Motivated by recent advancements in natural language processing (NLP), multi-modal AI and object detection, we use text-to-mask models and human feedback to learn compact and informative communication strategies that facilitate coordination among agents to improve the overall performance. We demonstrate the efficiency of our approach on various multi-agent tasks and provide insights into emergent communication behaviors observed during training.
Primary Area: reinforcement learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 13060
Loading