Research map / Vision-language and multimodal / Robot action and navigation
PixelPilot: Scalable Vision-Language-Action Models for End-to-End Autonomous Driving
Pin Tang, Guoqing Wang, Xiangxuan Ren, Zhongdao Wang, Guodongfang Zhao, Bailan Feng, Chao Ma
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 SFT to the base model (ablation) | Qwen2.5-VL-7B · nuScenes open-loop planning | average L2 (m): 1.92 -> 0.46 | Improved |
| add RL to the base model (ablation) | Qwen2.5-VL-7B · nuScenes open-loop planning | average L2 (m): 1.92 -> 0.49 | Improved |
| add RL after SFT (full two-stage) (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.46 -> 0.30 | Improved |
| add ego-centric consistent multi-view input (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.43 -> 0.30 | Improved |
| use multi-view ego-centric input instead of front view only (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.40 -> 0.30 | Improved |
| add perception-to-planning multi-task SFT (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.45 -> 0.30 | Improved |
| add trajectory reward (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.44 -> 0.30 | Improved |
| add perception IoU reward (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.41 -> 0.30 | Improved |
| add action reward (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.40 -> 0.30 | Improved |
| add fine-grained perception-to-planning CoT instead of none (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.41 -> 0.30 | Improved |
| use fine CoT instead of coarse CoT (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.35 -> 0.30 | Improved |
| include Waymo data in training (ablation) | PixelPilot · nuScenes open-loop planning | average L2 (m): 0.06 higher L2 without Waymo | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- AutoVLA — AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025) · better on 3 of 3 shared results
- EMMA — EMMA: Empowering Multi-modal Mamba with Structural and Hierarchical Alignment (ICLR 2025) · better on 1 of 2 shared results
- Orion — ORION: A Holistic End-to-End Autonomous Driving Framework by Vision-Language Instructed Action Generation (ICCV 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Driving (NeurIPS 2026 · similarity 0.92)
- FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving (NeurIPS 2025 · similarity 0.91)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.91)
- SpaceDrive: Infusing Spatial Awareness into VLM-based Autonomous Driving (CVPR 2026 · similarity 0.91)
- AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025 · similarity 0.91)
- LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving (NeurIPS 2026 · similarity 0.90)
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.
