行驶在寄存器之上
Driving on Registers
用相机感知的 register 词元把多相机 ViT 特征压成紧凑场景表示，再驱动轻量解码器生成并打分候选轨迹；这套「词元压缩 + 可解释子分数评分」的范式可直接迁移到其他需要多视角融合与实时推理的视觉任务。
原文依据 introduces camera-aware register tokens that compress multi-camera features into a compact scene representation
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论本文提出 DrivoR，一种用于端到端自动驾驶的简洁 Transformer 架构。该方法建立在预训练视觉 Transformer（ViT）之上，引入相机感知的寄存器词元，将多相机特征压缩为紧凑的场景表示，在不损失精度的前提下显著降低下游计算量。这些词元驱动两个轻量 Transformer 解码器，先产生候选轨迹，再对其打分。打分解码器通过模仿 oracle 进行学习，预测可解释的子分数（例如安全性或效率），从而在推理时实现行为条件化的驾驶。尽管设计极简，DrivoR 在 NAVSIM-v1/v2 以及闭环 HUGSIM 基准上均优于或持平强基线。结果表明，纯 Transformer 架构结合有针对性的词元压缩，足以实现准确、高效且自适应的端到端驾驶。代码与检查点见项目主页。
英文原文摘要 977 字符
We present DrivoR, a simple transformer-based architecture for end-to-end autonomous driving. Our approach builds on pretrained Vision Transformers (ViTs) and introduces camera-aware register tokens that compress multi-camera features into a compact scene representation, significantly reducing downstream computation without sacrificing accuracy. These tokens drive two lightweight transformer decoders that generate and then score candidate trajectories. The scoring decoder learns to mimic an oracle and predicts interpretable sub-scores e.g., safety or efficiency, enabling behavior-conditioned driving at inference. Despite its minimal design, DrivoR outperforms or matches strong baselines across NAVSIM-v1/v2, and closed-loop HUGSIM benchmarks. Our results show that a pure-transformer architecture, combined with targeted token compression, is sufficient for accurate, efficient, and adaptive end-to-end driving. Code and checkpoints are available via the project page.