arXiv cs.AI· Wei Wang (Robostreet research team Cambridge, MA, USA), Yiru Veronika Wang (Robostreet research team Cambridge, MA, USA), Yihan Wang (Robostreet research team Cambridge, MA, USA), Zihao Li (Robostreet research team Cambridge, MA, USA), Xiaohui Liang (Robostreet research team Cambridge, MA, USA)·· 12 小时前AI 评分38
DriveWorld3D：证据约束的 3D 世界建模与真实自动驾驶生成式仿真
DriveWorld3D Evidence Constrained 3D World Modeling and Generative Simulation for Real-World Autonomous Driving
AI 导读
DriveWorld3D 框架将时序视频学习、多传感器测量约束与离线生成式仿真结合，在线部分维护包含路面、体占据、动态实体与道路关系的统一世界状态，并区分物理占据、可观测性与证据来源。
正文 · 原文
Abstract:Three dimensional modeling for autonomous driving must continuously estimate road geometry, obstacle occupancy, traffic participant behavior, and observational uncertainty under limited computational resources. Existing reconstruction methods can synthesize realistic novel views, but visual fidelity does not sufficiently characterize the geometric reliability required for driving. Generative world models can expand scenarios and predict future observations, yet may introduce content unsupported by measurements. We propose DriveWorld3D, a framework combining temporal video learning, measurement constraints from multiple sensors, and offline generative simulation. Its online component maintains a unified world state comprising road surfaces, volumetric occupancy, dynamic entities, and road relations, while separately representing physical occupancy, observability, and evidence provenance. Action conditioned future prediction and risk sensitive planning connect this representation to driving decisions. Its offline component reconstructs recorded scenes using neural radiance fields or 3D Gaussian Splatting, then expands long tail conditions using generative models while restricting the use of generated content as geometric ground truth. We specify data partitioning, sensor ablations, fault injection, paired closed loop testing, and real to simulation consistency assessment. The central hypothesis is that explicit evidence management can reduce dangerous false free space predictions and occlusion risk within a matched computational budget while retaining acceptable driving efficiency. All performance claims remain subject to experimental validation. Keywords: autonomous driving; 3D occupancy; sensor fusion; uncertainty; generative world models; NeRF; 3D Gaussian Splatting; closed loop simulation
| Comments: | 10 pages |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| MSC classes: | 14J60 (Primary) |
| ACM classes: | I.3.7 |
| Cite as: | arXiv:2610.22450 [cs.RO] |
| (or arXiv:2610.22450v1 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22450 arXiv-issued DOI via DataCite (pending registration) |
|
| Journal reference: | Copilot4D: Learning Unsupervised World Models for Autonomous Driving via Discrete Diffusion. arXiv:2311.01017, 2023 |
Submission history
From: Wei Wang [view email]
[v1]
Thu, 15 Oct 2026 15:48:11 UTC (1,232 KB)
来源：arXiv cs.AI · arxiv.org