Research map / Multi-agent systems and games / Cooperative multi-agent RL
Learning to Simulate Self-driven Particles System with Coordinated Policy Optimization
Zhenghao Peng, Quanyi Li, Ka Ming Hui, Chunxiao Liu, Bolei Zhou
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 CoPO's local and global coordination instead of IPO | traffic MARL populations · Roundabout, Intersection, Tollgate, Bottleneck, and Parking Lot | success rate: 73.67 vs 70.81; 78.97 vs 60.47; 86.13 vs 82.90; 79.68 vs 72.43; 65.04 vs 61.05 | Improved |
| replace the learned LCF distribution with fixed N(0, 0.1²) (ablation) | CoPO · Roundabout, Intersection, Tollgate, Bottleneck, and Parking Lot | success rate: 62.38 vs 73.67; 70.23 vs 78.97; 60.47 vs 86.13; 71.29 vs 79.68; 59.14 vs 65.04 | Hurt |
| replace CoPO with IPO maximizing global reward directly (ablation) | CoPO · Roundabout, Intersection, Tollgate, Bottleneck, and Parking Lot | success rate: 0.00 { $\pm$ 0.00} on all five environments | Hurt |
| optimize neighborhood reward alone instead of balancing individual and neighborhood rewards (ablation) | CoPO · Roundabout, Intersection, Tollgate, Bottleneck, and Parking Lot | success rate: 65.70 vs 73.67; 66.83 vs 78.97; 57.30 vs 86.13; 71.62 vs 79.68; 6.34 vs 65.04 | Hurt |
| add centralized critics to CoPO (ablation) | CoPO · traffic environments | success rate | Hurt |
| use neighborhood mean-field inputs rather than concatenated nearby-agent states in centralized critics (ablation) | MFPO · Roundabout | training stability | Improved |
| add neighbors' actions alongside observations to centralized critics (ablation) | centralized critic variants · Roundabout | training stability | Improved |
| train with a distributional LCF rather than one fixed social-orientation factor | CoPO · traffic populations with heterogeneous IDM policies | success rate: IPO (25% IDM): 55.17; CoPO (25% IDM): 71.55; IPO (50% IDM): 52.14; CoPO (50% IDM): 63.34 | Improved |
Closest papers by meaning
- Learning Decentralized Partially Observable Mean Field Control for Artificial Collective Behavior (ICLR 2024 · similarity 0.87)
- Scalable Constrained Policy Optimization for Safe Multi-agent Reinforcement Learning (NeurIPS 2024 · similarity 0.86)
- Foundations of Multi-Agent Learning in Dynamic Environments: Where Reinforcement Learning Meets Strategic Decision-Making (AAAI 2025 · similarity 0.86)
- R3DM: Enabling Role Discovery and Diversity Through Dynamics Models in Multi-agent Reinforcement Learning (ICML 2025 · similarity 0.86)
- POGEMA: A Benchmark Platform for Cooperative Multi-Agent Pathfinding (ICLR 2025 · similarity 0.86)
- Controlling Behavioral Diversity in Multi-Agent Reinforcement Learning (ICML 2024 · 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.
