Research map / Multi-agent systems and games / Cooperative multi-agent RL
Beyond Rule-Based Agents: Active Markov Games for Realistic Multi-Agent Interaction in Autonomous Driving
Yuan Gui, Hongchen Luo, Jiao Wang, Liqi Qu
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 |
|---|---|---|---|
| train against a diverse opponent policy pool instead of rule-based opponents | single-agent PPO baseline · CARLA unprotected intersections | collision rate: 0.08 -> 0.60 | Hurt |
| add the opponent strategy pool (ablation) | full AMG co-evolution method · CARLA (multi-agent) | OOD success rate: 0.81 -> 0.98 | Improved |
| add the strategy inference module (ablation) | full AMG co-evolution method · CARLA (multi-agent) | OOD success rate: 0.84 -> 0.98 | Improved |
| add the map module (ablation) | full AMG co-evolution method · CARLA (multi-agent) | OOD success rate: 0.62 -> 0.98 | Improved |
| add the perfect perception module (ablation) | full AMG co-evolution method · CARLA (multi-agent) | OOD success rate: 0.54 -> 0.98 | Improved |
| add the strategy pool module (ablation) | full AMG co-evolution method · CARLA (multi-agent) | collision rate: 0.17 -> 0.02 | Improved |
| interactive co-evolution training instead of single-agent or rule-based training | ego agent · CARLA strategy diversity | strategy diversity: 0.3278 -> 0.5841 | Improved |
| increase TTC potential intensity beta from 0.5 to 0.9 (ablation) | potential-field reward · CARLA hyperparameter sensitivity | average return: 0.4: 118.3134, 0.6: 128.0711, 0.9: 106.6458 | Hurt |
| increase interaction weight to 0.7 (ablation) | potential-field reward · CARLA hyperparameter sensitivity | average return vs safety factor: return 135.3574 -> 128.1268; safety factor 3.78 -> 3.84 | Mixed |
| multi-agent IPPO baseline vs proposed method under diverse opponents | IPPO · CARLA unprotected intersections (diverse opponent) | collision rate: 0.29 -> 0.02 | Improved |
Closest papers by meaning
- Learning to Drive in New Cities Without Human Demonstrations (NeurIPS 2026 · similarity 0.88)
- MAPLE: Latent Multi-Agent Play for End-to-End Autonomous Driving (NeurIPS 2026 · similarity 0.88)
- Beyond Self-Play and Scale: A Behavior Benchmark for Generalization in Autonomous Driving (NeurIPS 2026 · similarity 0.88)
- World Models as Adversaries: Multi-Agent Self-Play Fine-Tuning for Robust Motion Planning (NeurIPS 2026 · similarity 0.88)
- IntersectionZoo: Eco-driving for Benchmarking Multi-Agent Contextual Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Robust Autonomy Emerges from Self-Play (ICML 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.
