Research map / Multi-agent systems and games / Cooperative multi-agent RL
Deep Reinforcement Learning for Communication Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
A doctoral-consortium style overview of the author's research on applying (multi-agent) RL to congestion control and to message dissemination in group communication networks, using graph convolutional RL.
Problem
Learning communication protocols and strategies in point-to-point congestion control and group communication (message dissemination in broadcast/dynamic networks) where heuristics require careful tuning.
Approach
Frames message dissemination in group communication networks as a game-theoretic MARL problem inspired by OLSR broadcast mechanisms, and proposes graph-convolution-based MARL methods (L-DGN, HL-DGN) to learn what and when to communicate; also applies RL to congestion control under dynamic link properties.
Key results
- Learned MARL strategy on static graphs outperforms a widely used heuristic (OLSR) for message dissemination, per the author's prior work cited in the summary.
- RL congestion control agents yield competitive results in networks with dynamic link properties.
Limitations
- Message dissemination results only evaluated on static graphs; dynamic graph structures are future work.
- No real-world protocol evaluation yet.
- Results are reported only qualitatively in this extended abstract.
Experimental setup
Baselines: OLSR heuristic CommNet ATOC
Builds on: CommNet ATOC Graph Convolutional Reinforcement Learning OLSR
Closest papers by meaning
- Bridging Training and Execution via Dynamic Directed Graph-Based Communication in Cooperative Multi-Agent Systems (AAAI 2025 · similarity 0.88)
- Exponential Topology-enabled Scalable Communication in Multi-agent Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Social Influence as Intrinsic Motivation for Multi-Agent Deep Reinforcement Learning (ICML 2019 · similarity 0.87)
- Multi-Agent Incentive Communication via Decentralized Teammate Modeling (AAAI 2022 · similarity 0.87)
- Learning Efficient and Robust Multi-Agent Communication via Graph Information Bottleneck (AAAI 2024 · similarity 0.86)
- Multi-Agent Reinforcement Learning with Communication-Constrained Priors (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.