Terrain Consistent Reference-Guided RL for Humanoid Navigation Autonomy
William D. Compton, Zachary Olkin, Aaron D. Ames · 2026
Read Terrain Consistent Reference-Guided RL for Humanoid Navigation Autonomy by William D. Compton, Zachary Olkin, Aaron D. Ames (2026) 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-2605-15517.txt.
Description
William D. Compton, Zachary Olkin, and Aaron D. Ames train perceptive locomotion policies for humanoid robots whose reference paths are adjusted during training to match terrain geometry. Footsteps are projected onto valid footholds, and swing-foot and center-of-mass paths are reshaped accordingly, exposing a clean velocity interface for standard navigation planners. In simulation the terrain-aware references track better, and on a Unitree G1 robot the system navigated over 70 meters outdoors through rough ground and stair flights with all sensing onboard.
Catalog record
| CREATOR | William D. Compton, Zachary Olkin, Aaron D. Ames |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38188 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.15517