Learning Perceptive Humanoid Locomotion over Challenging Terrain
Wandong Sun, Baoshi Cao, Long Chen, Yongbo Su, Yang Liu, Zongwu Xie, Hong Liu · 2025
Read Learning Perceptive Humanoid Locomotion over Challenging Terrain by Wandong Sun, Baoshi Cao, Long Chen, Yongbo Su, Yang Liu, Zongwu Xie, Hong Liu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-00692.txt.
Description
Wandong Sun and co-authors teach a humanoid robot to walk rough ground with the help of a teacher that sees clearly. Controllers that rely on body senses alone turn shaky on hard terrain, while height maps allow gait plans but arrive noisy in the real world. The team pairs the two in a distillation frame.
In their 2025 paper, an oracle policy with clean data sets the gold standard. The student copies the teacher while it also trains a world model that cleans sensor noise and gauges state through an information bottleneck. Scores jump where terrain reads are unreliable. Field tests in city streets and off-road ground ran for kilometers with no outside help.
Catalog record
| CREATOR | Wandong Sun, Baoshi Cao, Long Chen, Yongbo Su, Yang Liu, Zongwu Xie, Hong Liu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-00692