arXiv cs.AI· Dujun Nie, Ruiqi Song, Zixuan Wang, Hanqing Wang, Yifan Han, Baiyong Ding, Wenzhao Zheng, Hangbin Wu, Long Chen·· 4 小时前AI 评分47
OpenNWM：基于潜在动作预训练的通用导航世界模型
OpenNWM: A Generalist Navigation World Model with Latent Action Pretraining
AI 导读
研究者提出 OpenNWM，一种通过潜在动作预训练构建的通用导航世界模型，从无动作标注的大规模导航视频中学习通用潜在动作，再通过动作条件后训练与物理运动对齐。团队还构建了 NavAnywhere 数据集，包含来自 15 个异构来源的 1750 万条视觉观测，覆盖地面机器人、人类和无人机的五类场景。实验显示该模型在域内和域外基准上的视觉预测与导航性能均有提升。
正文 · 原文
Abstract:Navigation world models predict future observations conditioned on navigation actions to support planning, but their generalization across real-world environments remains constrained by the limited coverage of action-labeled data. Action-free navigation videos provide diverse visual experience, yet lack the recorded actions required for action-conditioned training. We introduce OpenNWM, a generalist navigation world model with latent action pretraining. It learns universal latent actions from large-scale action-free videos through visual prediction pretraining, and further aligns them with physical motions via action-conditioned post-training. To support video pretraining, we curate NavAnywhere, a large-scale action-free navigation video dataset comprising 17.5 million visual observations from 15 heterogeneous sources. It captures navigation by ground robots, humans, and drones, with diverse viewpoints and motion patterns across five broad scene types: residential areas, public and commercial spaces, urban settings, parks, and natural landscapes. Experiments demonstrate improved visual prediction on in-domain and out-of-domain benchmarks, together with improved navigation performance. These results show that latent-action pretraining enables diverse, action-free videos to improve dynamics learning and generalization for open-world navigation.
| Comments: | Project page: this https URL Code: this https URL |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.01844 [cs.RO] |
| (or arXiv:2611.01844v2 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2611.01844 arXiv-issued DOI via DataCite |
Submission history
From: Dujun Nie [view email]
[v1]
Sun, 1 Nov 2026 12:23:38 UTC (18,194 KB)
[v2]
Thu, 5 Nov 2026 04:49:34 UTC (18,194 KB)
来源：arXiv cs.AI · arxiv.org