Research map / Label-efficient and robust learning / Recognition and long-tail
ActiveAD: Planning-Oriented Active Learning for End-to-End Autonomous Driving
Han Lu, Xiaosong Jia, Yichen Xie, Siyu Sun, Wenlong Liao, Xiaokang Yang, Junchi Yan
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 |
|---|---|---|---|
| use Ego-Diversity instead of random initial selection (10% data) (ablation) | ActiveAD · nuScenes, VAD-Tiny | Average collision (%): 0.67 -> 0.41 | Improved |
| add Soft Collision criterion to Displacement Error (ablation) | ActiveAD incremental selection · nuScenes, 30% data | Average collision (%): 0.35 -> 0.26 | Improved |
| use only Displacement Error for active selection (ablation) | ActiveAD · nuScenes | Collision rate / L2 at 30% vs 20% | Mixed |
| add Agent Uncertainty criterion to DE and SC (ablation) | ActiveAD · nuScenes, 30% data | Average L2 (m): 0.73 -> 0.68 | Improved |
| add Agent Uncertainty criterion to DE and SC (ablation) | ActiveAD · nuScenes, 30% data | Average collision (%): 0.26 -> 0.21 | Improved |
| apply incremental selection on random initialization (ablation) | random init baseline · nuScenes, 30% data | Average L2 (m): 0.78 -> 0.71 | Improved |
| lower gamma from 1 to 0.5 in Ego-Diversity (more weight on minority scenes) (ablation) | Ego-Diversity initialization · nuScenes, 10% initial data | L2 (m): 0.90 -> 0.83 | Improved |
| use gamma 0.8 rather than 0.5 (ablation) | Ego-Diversity initialization · nuScenes, 10% initial data | Collision rate (%): 0.41 vs 0.43 | Mixed |
| use ActiveAD 30% selection instead of full-data training | VAD-Tiny · nuScenes planning | Average L2 (m): 0.70 -> 0.68 | Improved |
| use ActiveAD with 30% data on CARLA instead of full data | VAD-Tiny · CARLA Town05 Short | Drive Score: 63.11 -> 63.24 | Improved |
| use general active learning baselines (Coreset, VAAL, KECOR, ActiveFT) instead of random | VAD-Tiny · nuScenes planning | Average L2 / collision | No clear effect |
| increase data from 30% to 40-50% | ActiveAD · nuScenes planning | planning performance | No clear effect |
Closest papers by meaning
- nuReasoning: A Reasoning-Centric Dataset and Benchmark for Long-Tail Autonomous Driving (NeurIPS 2026 · similarity 0.86)
- Uncertainty-Guided Never-Ending Learning to Drive (CVPR 2024 · similarity 0.86)
- Continuously Learning, Adapting, and Improving: A Dual-Process Approach to Autonomous Driving (NeurIPS 2024 · similarity 0.86)
- Generalized Predictive Model for Autonomous Driving (CVPR 2024 · similarity 0.86)
- AlignDrive: Aligned Lateral-Longitudinal Planning for End-to-End Autonomous Driving (NeurIPS 2026 · similarity 0.86)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · 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.
