SafeDrive: Fine-Grained Safety Reasoning for End-to-End Driving in a Sparse World
It addresses safety for autonomous driving systems, representing an incremental improvement by integrating interpretable reasoning into existing end-to-end paradigms.
The paper tackles the challenge of ensuring safety in end-to-end driving by proposing SafeDrive, a framework that performs explicit safety reasoning through a trajectory-conditioned sparse world model, achieving state-of-the-art performance with metrics like 0.5% collision rate on NAVSIM and 66.8% driving score on Bench2Drive.
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.