Research map / Multi-agent systems and games / Cooperative multi-agent RL
Locality Matters: A Scalable Value Decomposition Approach for Cooperative Multi-Agent Reinforcement Learning
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 |
|---|---|---|---|
| Use LOMAQ with local rewards instead of the compared cooperative methods using global rewards | Cooperative MARL · Coupled-Multi-Cart-Pole | test return and convergence speed | Helped |
| Use LOMAQ with local rewards instead of the compared cooperative methods using global rewards | Cooperative MARL · Bounded-Cooperative-Navigation | test return and convergence speed | Helped |
| Learn local rewards from the global reward signal | LOMAQ · Coupled-Multi-Cart-Pole | test return | No clear effect |
| Use local rewards while acting greedily with independent Q-learning | IQL · Coupled-Multi-Cart-Pole and Bounded-Cooperative-Navigation | test return | Hurt |
| Use a singleton-agent partition rather than a joint partition (ablation) | LOMAQ · Matrix game representability experiment | representability | Hurt |
| Increase the neighborhood radius κ (ablation) | LOMAQ · Coupled-Multi-Cart-Pole | convergence speed | Hurt |
| Increase the neighborhood radius κ (ablation) | LOMAQ · Coupled-Multi-Cart-Pole | converged policy quality | Helped |
| Allow reward components to depend on groups of agents, without regularization (ablation) | Reward decomposition from a global signal · Bounded-Cooperative-Navigation, two agents and one landmark | local reward information | Hurt |
| Add cardinality-dependent regularization to the group reward decomposition (ablation) | Reward decomposition from a global signal · Bounded-Cooperative-Navigation, two agents and one landmark | local reward information | Helped |
| Restrict the reward decomposition to individual agents when rewards depend on a shared landmark (ablation) | Reward decomposition from a global signal · Bounded-Cooperative-Navigation, two agents and one landmark | global reward approximation | Hurt |
Closest papers by meaning
- Revisiting Some Common Practices in Cooperative Multi-Agent Reinforcement Learning (ICML 2022 · similarity 0.92)
- Gradient-Protected Value Decomposition for Cooperative Multi-Agent Reinforcement Learning (AAAI 2026 · similarity 0.91)
- Q-value Path Decomposition for Deep Multiagent Reinforcement Learning (ICML 2020 · similarity 0.90)
- Benchmarking Multi-Agent Deep Reinforcement Learning Algorithms in Cooperative Tasks (NeurIPS 2021 · similarity 0.90)
- Optimistic Value Instructors for Cooperative Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.90)
- SMIX(λ): Enhancing Centralized Value Functions for Cooperative Multi-Agent Reinforcement Learning (AAAI 2020 · similarity 0.90)
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.