Research map / Label-efficient and robust learning / Domain adaptation and generalization
RoCA: Robust Cross-Domain End-to-End Autonomous Driving
Rajeev Yasarla, Shizhong Han, Hsin-Pai Cheng, Apratim Bhattacharyya, Shweta Mahajan, Litian Liu, Yunxiao Shi, Risheek Garrepalli, Hong Cai, Fatih Porikli
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 RoCA GP regularization during source training | ORION · Bench2Drive closed-loop | Driving Score: 77.74 -> 80.38 | Improved |
| add RoCA GP regularization during source training | SSR · Bench2Drive closed-loop | Mean ability: 35.38 -> 48.98 | Improved |
| add RoCA regularization, zero-shot sim-to-real | ORION · nuScenes full val (zero-shot) | Avg. L2: 0.98 -> 0.68 | Improved |
| use RoCA GT adaptation instead of direct finetuning | SparseDrive-S · nuScenes Boston to Singapore | Avg. L2: 0.55 -> 0.49 | Improved |
| use RoCA unsupervised adaptation instead of GT finetuning (ablation) | SparseDrive-S · nuScenes Boston to Singapore | Avg. L2: 0.55 -> 0.71 | Hurt |
| add RoCA regularization only (no adaptation) (ablation) | VAD-Tiny · nuScenes Boston to Singapore | Avg. Col. (%): 0.43 -> 0.21 | Improved |
| select active-learning samples by GP variance instead of random (ablation) | RoCA adaptation · Singapore to Boston 15% | Avg. L2: 0.552 -> 0.513 | Improved |
| use GP uncertainty instead of Deep Ensemble for sampling (ablation) | RoCA · S to B 15% | Avg. L2: 0.539 -> 0.513 | Improved |
| add RoCA unsupervised adaptation to degraded-image model (ablation) | SparseDrive-S · nuScenes snow (full val) | Avg. L2: 0.857 -> 0.628 | Improved |
| add RoCA regularization (unsupervised adaptation) on NAVSIM | Orion · NAVSIM v1 | PDMS: 84.4 -> 87.9 | Improved |
| add RoCA on DriveArena zero-shot | Orion · DriveArena | PDMS: 71.6 -> 91.2 | Improved |
| add RoCA training regularization (ablation) | SparseDrive-S · Low-light degradation (full val) | Avg. Col. (%): 0.145 -> 0.129 | Improved |
Closest papers by meaning
- Model-Based Policy Adaptation for Closed-Loop End-to-end Autonomous Driving (NeurIPS 2025 · similarity 0.89)
- DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving (ICLR 2025 · similarity 0.87)
- CoC-VLA: Delving into Adversarial Domain Transfer for Explainable Autonomous Driving via Chain-of-Causality Visual-Language-Action Model (NeurIPS 2025 · similarity 0.87)
- Bench2Drive: Towards Multi-Ability Benchmarking of Closed-Loop End-To-End Autonomous Driving (NeurIPS 2024 · similarity 0.87)
- Less Language, More Latents: Annotation-Efficient VLAs for Driving (NeurIPS 2026 · similarity 0.87)
- Distilling Multi-modal Large Language Models for Autonomous Driving (CVPR 2025 · 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.
