arXiv cs.AI· Haoran Xu, Peixi Peng, Shijun Guo, Guang Tan, Yiqian Chang, Yisen Zhao, Shuaixian Wang, Yonghong Tian·· 5 小时前AI 评分29
Aligned Delta-Triplane Transformer：面向自动驾驶的占用世界模型
Aligned Delta-Triplane Transformers as Occupancy World Models
AI 导读
研究者提出 Aligned Delta-Triplane Transformer（ADTT），一种紧凑的 4D 占用世界模型（OWM），用于自动驾驶场景预测。该方法通过自车运动补偿显式对齐历史 triplane，仅预测未来场景变化，并借助自车运动查询与可学习外部查询由 query-conditioned Transformer 生成残差。
正文 · 原文
Abstract:Occupancy World Models (OWMs) aim to predict future 3D occupancy scenes from historical observations and future ego motions, providing a useful world simulation tool for autonomous driving. Existing methods usually rely on large networks to implicitly align multi-frame historical states and predict future occupancy in a full-state manner, which is costly and redundant because most scene regions remain unchanged over short time intervals. In this paper, we propose Aligned Delta-Triplane Transformer (ADTT), a compact 4D OWM that explicitly aligns historical triplanes with ego-motion compensation and predicts only future scene changes. Based on the aligned triplane prior, a query-conditioned Transformer predicts residual changes, where ego-motion queries guide controllable future motion and learnable external queries capture scene changes beyond ego motion. The predicted changes are added to the aligned prior and decoded into future occupancy scenes. Experiments on Occ3D-nus show that ADTT achieves state-of-the-art forecasting performance with fewer parameters and higher FPS. Our code is publicly available online: this https URL.
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2503.07338 [cs.RO] |
| (or arXiv:2503.07338v5 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2503.07338 arXiv-issued DOI via DataCite |
Submission history
From: Haoran Xu [view email]
[v1]
Mon, 10 Mar 2025 13:50:23 UTC (19,127 KB)
[v2]
Thu, 15 May 2025 08:04:19 UTC (19,468 KB)
[v3]
Sat, 27 Sep 2025 18:26:34 UTC (6,300 KB)
[v4]
Sat, 29 Nov 2025 15:14:29 UTC (7,724 KB)
[v5]
Thu, 29 Oct 2026 07:49:44 UTC (3,089 KB)
来源：arXiv cs.AI · arxiv.org