# Drive3R：面向自动驾驶的块式流式几何基础模型

> 原文链接：https://robotics.hellobike.com/papers/2026-10-10-2610-15112
> 频道：papers | 更新日期：2026-10-10 | 标签：arxiv, cs.RO
> 核心摘要：针对长时程驾驶序列稠密多视角3D重建难扩展，Drive3R以分块流式注意力和重锚定，在nuScenes、KITTI、Waymo验证。

# Drive3R：面向自动驾驶的块式流式几何基础模型

::: v-pre
> 原文链接：[Drive3R: Block-wise Streaming Geometric Foundation Models for Autonomous Driving](https://arxiv.org/abs/2610.15112)（arXiv 2610.15112 · 发布日期：2026-10-10）

**Drive3R: Block-wise Streaming Geometric Foundation Models for Autonomous Driving**

针对长时程驾驶序列稠密多视角3D重建难扩展，Drive3R以分块流式注意力和重锚定，在nuScenes、KITTI、Waymo验证。

## 核心要点

- 高保真闭环仿真依赖稠密多视角3D重建，但长时程序列受显存和全局注意力二次复杂度限制。
- Drive3R采用块式因果注意力，缓解全双向注意力扩展瓶颈，并保留局部未来上下文。
- 置信度感知重锚定检测不可靠流式状态，重置KV缓存，并用SE(3)对齐拼接刷新片段。
- 实验显示其在nuScenes、KITTI、Waymo达新SOTA；nuScenes多项指标优于DVGT-2。

**验证范围与局限**：验证范围为nuScenes、KITTI和Waymo上的自动驾驶序列实验。

## 英文摘要 (Abstract)

High-fidelity closed-loop simulation for autonomous driving increasingly relies on dense multi-view 3D reconstruction. While recent feed-forward geometric foundation models have revolutionized per-scene optimization, applying them to continuous, long-horizon driving sequences is computationally intractable due to GPU memory limits and the quadratic complexity of global attention. Existing scaling strategies are fundamentally limited, as causal streaming suffers from compounding scale drift and trajectory deviation, while rigid "chunk-then-align" heuristics fragment global context and produce trajectory discontinuities. To address these limitations, we introduce Drive3R, a block-wise streaming geometric foundation model. Its block-wise causal attention resolves the scalability bottleneck of fully bidirectional attention while avoiding the loss of local future context in purely causal streaming. Its confidence-aware re-anchoring further mitigates long-range drift by detecting unreliable streaming states, re-initializing the KV cache, and stitching refreshed segments through SE(3) alignment. Extensive experiments demonstrate that Drive3R establishes a new state-of-the-art on nuScenes, KITTI, and Waymo. Specifically, on nuScenes, it improves over the recent DVGT-2 baseline from 0.1806 to 0.1081 in AbsRel, from 0.6750 to 0.5344 in point-cloud Accuracy, and from 1.5812 to 0.9092 in ATE.

作者：Tianyi Yang, Haojun Yang, Jipeng Xiao, Xuesong Chen, Shaoshuai Shi, Li Jiang, Zhipeng Zhang
:::

## 原始参考来源 (Sources)

- [arXiv 2610.15112：Drive3R: Block-wise Streaming Geometric Foundation Models for Autonomous Driving](https://arxiv.org/abs/2610.15112)