Research map / Multi-agent systems and games / Cooperative multi-agent RL
Learning to Schedule Communication in Multi-agent Reinforcement Learning
Daewoo Kim, Sangwoo Moon, David Hostallero, Wan Ju Kang, Taeyoung Lee, Kyunghwan Son, Yung Yi
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| use learned Top(1) scheduling instead of round-robin scheduling | SchedNet · Predator and Prey (PP) | task performance: 43% | Improved |
| use learned scheduling instead of round-robin scheduling | SchedNet · Cooperative Communication and Navigation (CCN) | time to complete task: 32% | Improved |
| use learned Top(1) scheduling rather than Softmax(1) scheduling | SchedNet · Predator and Prey (PP) | time to capture prey | Improved |
| use learned Top(1) scheduling rather than Softmax(1) scheduling | SchedNet · Cooperative Communication and Navigation (CCN) | time to complete task | Improved |
| add communication with learned scheduling | SchedNet · Predator and Prey (PP) | time to capture prey | Improved |
| add communication with learned scheduling | SchedNet · Cooperative Communication and Navigation (CCN) | time to complete task | Improved |
| train the encoder jointly with the scheduler and action network instead of fixing a pre-trained auto-encoder (ablation) | SchedNet-Top(1) · Predator and Prey (PP) | time to complete task, normalized to FC: 3.408 -> 2.030 | Improved |
| use communication with learned scheduling rather than no communication (ablation) | SchedNet · Predator and Prey (PP) | time to capture prey | Improved |
| increase message bandwidth (ablation) | SchedNet · Predator and Prey (PP) | time to complete task | Improved |
| increase the number of agents allowed to broadcast (ablation) | SchedNet · Predator and Prey (PP) | time to complete task | Improved |
| add a GRU layer to the encoder, action selector, and weight generator (ablation) | SchedNet · Predator and Prey (PP) | average steps to complete task | No clear effect |
| add a GRU layer to the network (ablation) | IDQN · Predator and Prey (PP) | average steps to complete task | Improved |
Closest papers by meaning
- Learning Multi-Agent Communication through Structured Attentive Reasoning (NeurIPS 2020 · similarity 0.90)
- Learning Efficient Multi-agent Communication: An Information Bottleneck Approach (ICML 2020 · similarity 0.89)
- Multi-Agent Reinforcement Learning with Communication-Constrained Priors (NeurIPS 2025 · similarity 0.88)
- Efficient Communication in Multi-Agent Reinforcement Learning with Implicit Consensus Generation (AAAI 2025 · similarity 0.88)
- Bridging Training and Execution via Dynamic Directed Graph-Based Communication in Cooperative Multi-Agent Systems (AAAI 2025 · similarity 0.87)
- Learning Multi-Agent Communication from Graph Modeling Perspective (ICLR 2024 · similarity 0.87)
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.
