Research map / Multi-agent systems and games / Cooperative multi-agent RL
RODE: Learning Roles to Decompose Multi-Agent Tasks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Decompose action spaces into restricted roles using action-effect clusters | RODE · SMAC 14 scenarios | scenarios with highest median test win rate by at least 1/32: 10 of all 14 scenarios | Helped |
| Transfer the learned policy to unseen maps with more agents and enemies, without further policy training | RODE trained on corridor · Unseen maps transferred from corridor | win rate: 50% | Helped |
| Use conventional Q-networks instead of action representations in policy learning (ablation) | RODE · SMAC three most difficult scenarios | performance | Hurt |
| Allow role policies to use the full action space and use conventional Q-networks (ablation) | RODE · SMAC three most difficult scenarios | performance | Hurt |
| Use random restricted role action spaces instead of effect-based clusters (ablation) | RODE · SMAC three most difficult scenarios | performance | Hurt |
| Remove action-space restrictions while retaining the hierarchical structure (ablation) | RODE · SMAC three most difficult scenarios | performance | No clear effect |
| Replace the recurrent role selector with a fully connected selector conditioned on local observations (ablation) | RODE · SMAC several environments | performance | Hurt |
| Vary the role interval (ablation) | RODE · SMAC several environments | performance | Mixed |
| Use RODE on easy maps where substantial exploration is unnecessary | RODE · SMAC easy maps | samples to achieve similar performance | Hurt |
Closest papers by meaning
- Effective and Stable Role-Based Multi-Agent Collaboration by Structural Information Principles (AAAI 2023 · similarity 0.91)
- ROMA: Multi-Agent Reinforcement Learning with Emergent Roles (ICML 2020 · similarity 0.89)
- Attention-Guided Contrastive Role Representations for Multi-agent Reinforcement Learning (ICLR 2024 · similarity 0.88)
- R3DM: Enabling Role Discovery and Diversity Through Dynamics Models in Multi-agent Reinforcement Learning (ICML 2025 · similarity 0.88)
- Learning Generalizable Skills from Offline Multi-Task Data for Multi-Agent Cooperation (ICLR 2025 · similarity 0.86)
- Trajectory-Class-Aware Multi-Agent Reinforcement Learning (ICLR 2025 · similarity 0.84)
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.