Research map / Reinforcement learning / Visual and representation RL
Learning Situational Driving
Check 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 three experts instead of one policy component (ablation) | Nav. Dynamic policy · CARLA Town 2, Nav. Dynamic | success rate (%): 78 -> 92 | Helped |
| Add dynamic-obstacle examples to training (ablation) | Monolithic behavior-cloning policy · CARLA Town 2, Navigation | success rate (%): 96 -> 78 | Hurt |
| Add dynamic-obstacle examples to training (ablation) | Monolithic behavior-cloning policy · CARLA Town 2, Nav. Dynamic | success rate (%): 40 -> 78 | Helped |
| Use three shared-backbone experts rather than one monolithic policy (ablation) | CARLA driving policy · CARLA Town 2, Nav. Dynamic | success rate (%): 75 -> 89 | Helped |
| Use separate expert backbones rather than a shared backbone (ablation) | Three-expert MoE · CARLA Town 2, Nav. Dynamic | success rate (%): 89 -> 94 | Helped |
| Increase shared-backbone expert count from three to five (ablation) | MoE-Branched · CARLA Town 2, Nav. Dynamic | success rate (%): 89 -> 90 | Helped |
| Increase shared-backbone expert count from five to eight (ablation) | MoE-Branched · CARLA Town 2, Nav. Dynamic | success rate (%): 90 -> 87 | Hurt |
| Increase separate-backbone expert count from three to five (ablation) | MoE · CARLA Town 2, Nav. Dynamic | success rate (%): 94 -> 93 | Hurt |
| Task-optimize the mixture coefficients and context projection (ablation) | MoE (K=3) · CARLA Town 2, Nav. Dynamic | success rate (%): 94 -> 98 | Helped |
| Use LSD+ rather than CILRS* | Driving policy · AnyWeather, New Town, Nav. Dynamic | success rate (%): 75.6 -> 78.4 | Helped |
| Use LSD+ rather than LBC | Driving policy · CARLA 0.8.4, New Town, Nav. Dynamic | success rate (%): 99 -> 98 | Hurt |
| Use LSD+ rather than Expert | Driving policy · NoCrash, New Town, Dense | success rate (%): 41 ± 2 -> 30 ± 4 | Hurt |
Closest papers by meaning
- SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving (CVPR 2026 · similarity 0.89)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.88)
- Hybrid-Driving: An Autonomous Driving Decision Framework Integrating Large Language Models, Knowledge Graphs and Driving Rules (AAAI 2025 · similarity 0.88)
- Continuously Learning, Adapting, and Improving: A Dual-Process Approach to Autonomous Driving (NeurIPS 2024 · similarity 0.88)
- DriveCombo: Benchmarking Compositional Traffic Rule Reasoning in Autonomous Driving (CVPR 2026 · similarity 0.87)
- DriveAgent-R1: Advancing VLM-based Autonomous Driving with Active Perception and Hybrid Thinking (ICLR 2026 · 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.