arXiv：cs.AI· Shaoting Zhu, Ziwen Zhuang, Mengjie Zhao, Kun-Ying Lee, Hang Zhao·· 18 小时前
Hiking in the Wild：面向人形机器人的可扩展感知跑酷框架
Hiking in the Wild: A Scalable Perceptive Parkour Framework for Humanoids
AI 导读
Hiking in the Wild 是一个可扩展的端到端感知式腿部跑酷框架，将深度输入与本体感知直接映射为关节动作，无需外部定位与建图。该框架引入 Terrain Edge Detection 结合 Foot Volume Points 的落脚安全机制，以及 Flat Patch Sampling 策略来缓解 reward hacking。
正文
Abstract:Achieving robust humanoid locomotion in complex, unstructured environments requires transitioning from reactive proprioception to proactive perception. However, integrating exteroception remains a significant challenge: existing approaches often fail to achieve high-speed, stable locomotion across diverse, unstructured terrains. In this work, we present Hiking in the Wild, an extensible, end-to-end perceptive legged parkour framework designed for robust humanoid hiking. Our approach maps depth inputs and proprioception directly to joint actions, without relying on external localization and mapping. To ensure safety and training stability, we introduce two key mechanisms: a foothold safety mechanism combining extensible Terrain Edge Detection with Foot Volume Points to reduce catastrophic slippage on edges, and a Flat Patch Sampling strategy that mitigates reward hacking by generating feasible navigation targets. Extensive experiments on the Unitree G1 humanoid robot demonstrate that our policy enables robust traversal of complex terrains at speeds up to 2.5 m/s.
| Comments: | Project Page: this https URL |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2601.07718 [cs.RO] |
| (or arXiv:2601.07718v2 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2601.07718 arXiv-issued DOI via DataCite |
Submission history
From: Shaoting Zhu [view email]
[v1]
Mon, 12 Jan 2026 16:50:50 UTC (7,917 KB)
[v2]
Tue, 13 Oct 2026 14:11:21 UTC (9,357 KB)
来源：arXiv：cs.AI · arxiv.org