SafeDrive: Fine-Grained Safety Reasoning for End-to-End Driving in a Sparse World

Abstract

The end-to-end (E2E) paradigm, which maps sensor inputs directly to driving decisions, has recently attracted significant attention due to its unified modeling capability and scalability. However, ensuring safety in this unified framework remains one of the most critical challenges. In this work, we propose SafeDrive, an E2E planning framework designed to perform explicit and interpretable safety reasoning through a trajectory-conditioned Sparse World Model. SafeDrive comprises two complementary networks: the Sparse World Network (SWNet) and the Fine-grained Reasoning Network (FRNet). SWNet constructs trajectory-conditioned sparse worlds that simulate the future behaviors of critical dynamic agents and road entities, providing interaction-centric representations for downstream reasoning. FRNet then evaluates agent-specific collision risks and temporal adherence to drivable regions, enabling precise identification of safety-critical events across future timesteps. SafeDrive achieves state-of-the-art performance on both open-loop and closed-loop benchmarks. On NAVSIM, it records a PDMS of 91.6 and an EPDMS of 87.5, with only 61 collisions out of 12,146 scenarios (0.5%). On Bench2Drive, SafeDrive attains a 66.8% driving score.

Method

Overall architecture of SafeDrive. ProposalNet evaluates the scene-level safety of anchor trajectories using BEV features and selects safety-aware candidates. SWNet constructs trajectory-conditioned Sparse Worlds by simulating the future behaviors of dynamic agents and road entities. FRNet performs fine-grained safety reasoning by estimating pair-wise No at-fault Collision score and evaluating Time-wise Drivable Area Compliance score over time, enabling interpretable and temporally grounded safety assessment.

Step-by-step Fine-grained Safety Prediction Process

Main Result

Qualitative Comparison with State-of-the-Art

Safety Reasoning in Open-loop

Pair-wise Collision Check

Time-wise Drivable Area Compliance Check

Safety Reasoning in Closed-loop

BibTeX

@inproceedings{safedrive,
  title={SafeDrive: Fine-Grained Safety Reasoning for End-to-End Driving in a Sparse World},
  author={Kim, Jungho and Oh, Jiyong and Yu, Seunghoon and Shin, Hongjae and Kwak, Donghyuk and Choi, Jun Won},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2026}
}