Research map / Multi-agent systems and games / Human-AI and goal-conditioned
ScenePilot: Controllable Boundary-Driven Critical Scenario Generation for Autonomous Driving
Qiyu Ruan, YUXUAN WANG, He Li, Zhenning Li, Cheng-zhong Xu
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 physical-feasibility constraint to AV-risk-only boundary (ablation) | ScenePilot · SafeBench Red-light Running | CR: 0.37 -> 0.93 | Improved |
| add AV-risk constraint to physical-feasibility-only boundary (ablation) | ScenePilot · SafeBench Red-light Running | CR: 0.75 -> 0.93 | Improved |
| add physical-feasibility constraint to AV-risk-only boundary (ablation) | ScenePilot · SafeBench Right Turn | CR: 0.12 -> 0.91 | Improved |
| add AV-risk constraint to physical-feasibility-only boundary (ablation) | ScenePilot · SafeBench Right Turn | CR: 0.74 -> 0.91 | Improved |
| use both boundaries rather than AV-risk only (ablation) | ScenePilot · SafeBench Red-light Running | Gap Coverage Score: 0.096 -> 0.209 | Improved |
| use both boundaries rather than physical-feasibility only (ablation) | ScenePilot · SafeBench Right Turn | Gap Coverage Score: 0.133 -> 0.351 | Improved |
| use ScenePilot instead of ChatScene as best baseline | scenario generator · SafeBench (8 base scenarios) | CR: 0.831 -> 0.893 | Improved |
| use ScenePilot instead of KING on heterogeneous AV stacks | scenario generator · SafeBench Scenario 6 | CR: 0.380 -> 0.490 | Improved |
| add 30 or 50 benign background vehicles | ScenePilot · SafeBench Scenario 6 | CR: 0.89 -> 0.91 | No clear effect |
| add 50 benign background vehicles | ScenePilot · SafeBench Scenario 6 | OS: 0.507 -> 0.492 | No clear effect |
| adversarially fine-tune ego on ScenePilot scenes | SAC ego policy · SafeBench held-out routes | CR: 0.854 -> 0.072 | Improved |
| fine-tune on ScenePilot scenes instead of ChatScene scenes | SAC ego policy · SafeBench held-out routes | OS: 0.869 -> 0.898 | Improved |
Closest papers by meaning
- Generating Useful Accident-Prone Driving Scenarios via a Learned Traffic Prior (CVPR 2022 · similarity 0.89)
- Steerable Adversarial Scenario Generation through Test-Time Preference Alignment (ICLR 2026 · similarity 0.89)
- DiffScene: Diffusion-Based Safety-Critical Scenario Generation for Autonomous Vehicles (AAAI 2025 · similarity 0.88)
- Scenario Dreamer: Vectorized Latent Diffusion for Generating Driving Simulation Environments (CVPR 2025 · similarity 0.88)
- Hybrid-Driving: An Autonomous Driving Decision Framework Integrating Large Language Models, Knowledge Graphs and Driving Rules (AAAI 2025 · similarity 0.87)
- Generating Traffic Scenarios via In-Context Learning to Learn Better Motion Planner (AAAI 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.
