Research map / Multi-agent systems and games / Cooperative multi-agent RL
Learning Efficient Multi-agent Cooperative Visual Exploration
Chao Yu, Xinyi Yang, Jiaxuan Gao, Huazhong Yang, Yu Wang, Yi Wu
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add policy distillation over per-scene teacher policies (ablation) | MAANS · Habitat Gibson train scenes, N=2 agents | Steps to 90% coverage: 180.25 -> 158.55 | Improved |
| add policy distillation over per-scene teacher policies (ablation) | MAANS · Habitat Gibson test scenes, N=2 agents | Steps to 90% coverage: 159.53 -> 144.16 | Improved |
| use MAANS instead of best planning baseline on train scenes | RRT · Habitat Gibson train scenes, N=2 agents | Steps to 90% coverage: 199.59 -> 158.55 | Improved |
| use MAANS instead of best planning baseline on test scenes | Voronoi · Habitat Gibson test scenes, N=2 agents | Steps to 90% coverage: 156.68 -> 144.16 | Improved |
| use MAANS instead of Voronoi for mutual overlap on test scenes | Voronoi · Habitat Gibson test scenes, N=2 agents | Mutual Overlap: 0.51 -> 0.54 | Hurt |
| replace multi-agent planners APF and WMA-RRT with single-agent RRT | APF/WMA-RRT · Habitat Gibson train scenes, N=2 agents | Steps to 90% coverage: 268.20 -> 199.59 | No clear effect |
| zero-shot transfer MAANS to N=3 on train scenes versus Voronoi | Voronoi · Habitat Gibson train, N=3 | Steps to 90% coverage: 155.13 -> 127.88 | Improved |
| zero-shot transfer MAANS to N=4 on train scenes versus RRT | RRT · Habitat Gibson train, N=4 | Steps to 90% coverage: 140.57 -> 114.75 | Improved |
| zero-shot transfer MAANS to N=4 on test scenes versus Voronoi | Voronoi · Habitat Gibson test, N=4 | Steps to 90% coverage: 101.90 -> 109.07 | Hurt |
| team size increases mid-episode (2 to 3) on train scenes | RRT · Habitat Gibson train, 2=>3 | Steps to 90% coverage: 185.94 -> 148.82 | Improved |
| team size decreases mid-episode (3 to 2) on train scenes | RRT · Habitat Gibson train, 3=>2 | Steps to 90% coverage: 187.93 -> 145.14 | Improved |
| add Spatial-TeamFormer instead of average pooling (ablation) | MSP · Gibson Colebrook and Dryville | Steps | Improved |
Closest papers by meaning
- MNE-SLAM: Multi-Agent Neural SLAM for Mobile Robots (CVPR 2025 · similarity 0.89)
- Reasoning in Visual Navigation of End-to-end Trained Agents: A Dynamical Systems Approach (CVPR 2025 · similarity 0.88)
- Imagine, Initialize, and Explore: An Effective Exploration Method in Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.87)
- GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- MAGiC-SLAM: Multi-Agent Gaussian Globally Consistent SLAM (CVPR 2025 · similarity 0.86)
- Acting Beyond Learning: Imagination-Assisted Decision-Making in the Visual-based Multi-Agent Cooperative Scenarios (AAAI 2025 · similarity 0.86)
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.
