Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Exploring the Limitations of Behavior Cloning for Autonomous Driving
Felipe Codevilla, Eder Santana, Antonio M. Lopez, Adrien Gaidon
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 speed prediction branch (ablation) | CILR · CARLA original, Nav. Dynamic (new town & weather) | success rate (%): 82 -> 90 | Improved |
| add speed prediction branch (ablation) | CILR · NoCrash Dense (new town & weather) | success rate (%): 13 -> 24 | Improved |
| add speed prediction branch (ablation) | CILR · NoCrash Dense (training conditions) | success rate (%): 28 -> 42 | Improved |
| add speed prediction branch (ablation) | CILR · NoCrash Dense (new weather) | success rate (%): 27 -> 47 | Improved |
| add speed prediction branch (ablation) | CILR · NoCrash Empty (new town & weather) | success rate (%): 66 -> 90 | Improved |
| initialize with ImageNet pretraining (ablation) | CILRS · NoCrash Empty | variance of success rate (%): 23% -> 4% | Improved |
| initialize with ImageNet pretraining (ablation) | CILRS · NoCrash Dense | variance of success rate (%): 42% -> 38% | Improved |
| increase traffic from empty to dense | CILRS · NoCrash (training / new conditions) | success rate: -55% / -66% | Hurt |
| change to new town and weather in empty town | CILRS · NoCrash Empty | success rate: -7% | Hurt |
| increase training data from 10 to 100 hours (ablation) | CILRS · NoCrash Dense (new town & weather) | success rate | No clear effect |
| increase amount of training data (ablation) | CILRS · NoCrash | episodes failed by inertia | Hurt |
| train with a different random seed | CILRS · NoCrash Dense (new weather & town) | success rate | Mixed |
Closest papers by meaning
- Feedback-Guided Autonomous Driving (CVPR 2024 · similarity 0.91)
- Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning (NeurIPS 2024 · similarity 0.90)
- Beyond Self-Play and Scale: A Behavior Benchmark for Generalization in Autonomous Driving (NeurIPS 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)
- DriveGPT: Scaling Autoregressive Behavior Models for Driving (ICML 2025 · similarity 0.88)
- ReSim: Reliable World Simulation for Autonomous Driving (NeurIPS 2025 · similarity 0.88)
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.
