Research map / Multi-agent systems and games / Cooperative multi-agent RL
Shield Decentralization for Safe Multi-Agent Reinforcement Learning
Daniel Melcer, Christopher Amato, Stavros Tripakis
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 |
|---|---|---|---|
| add decentralized shield instead of no shield | tabular Q-learning · Gridworld ISR fixed start | violations: 12.3 -> 0 | Improved |
| add decentralized shield instead of no shield | tabular Q-learning · Gridworld Pentagon fixed start | violations: 4.8 -> 0 | Improved |
| replace centralized shield with decentralized shield | tabular Q-learning · Gridworld ISR fixed start | evaluation return: 90.4 -> 89.9 | No clear effect |
| replace centralized shield with decentralized shield | tabular Q-learning · Gridworld Pentagon random start | evaluation return: 88.4 -> 80.3 | Hurt |
| replace centralized shield with decentralized shield | tabular Q-learning · Gridworld SUNY fixed start | evaluation return: 83.3 -> 86.6 | Improved |
| replace centralized shield with decentralized shield | DQN · Particle-Momentum partial observability, random start | evaluation return: 70.8 -> 80.5 | Improved |
| replace centralized shield with decentralized shield | DQN · Particle-Momentum full observability, fixed start | evaluation return: 91.6 -> 91.6 | No clear effect |
| add decentralized shield instead of no shield | DQN · Particle-Momentum partial observability, random start | evaluation return: 33.2 -> 80.5 | Improved |
| add decentralized shield instead of no shield | DQN · Particle-Momentum partial observability, random start | violations: 115.7 -> 0 | Improved |
| add decentralized shield instead of no shield | DQN · Particle-Momentum full observability, fixed start | violations: 3.0 -> 0 | Improved |
| decentralize the shield instead of centralized | shield synthesis · Gridworld collision | average safe joint actions: 23.30 -> 22.07 | No clear effect |
| decentralize the shield instead of centralized | shield synthesis · Particle-Momentum | average safe actions: 21.67 -> 20.81 | No clear effect |
Closest papers by meaning
- Tackling Safe and Efficient Multi-Agent Reinforcement Learning via Dynamic Shielding (Student Abstract) (AAAI 2023 · similarity 0.94)
- Safe Reinforcement Learning via Shielding under Partial Observability (AAAI 2023 · similarity 0.90)
- Probabilistic Shielding for Safe Reinforcement Learning (AAAI 2025 · similarity 0.87)
- Scalable Constrained Policy Optimization for Safe Multi-agent Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- HMARL-CBF – Hierarchical Multi-Agent Reinforcement Learning with Control Barrier Functions for Safety-Critical Autonomous Systems (NeurIPS 2025 · similarity 0.87)
- Explainably Safe Reinforcement Learning (NeurIPS 2025 · 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.
