From Simulation to Field: Learning Terrain Traversability for Real-World Deployment
Fetullah Atas, Grzegorz Cielniak, Lars Grimstad · 2025
Read From Simulation to Field: Learning Terrain Traversability for Real-World Deployment by Fetullah Atas, Grzegorz Cielniak, Lars Grimstad (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-2501-06904.txt.
Description
Fetullah Atas, Grzegorz Cielniak, and Lars Grimstad teach robots to judge which ground they can cross in wild outdoor terrain. Judging what can be crossed in a forest, given slopes, bumps, and rocks, is central to self-driving navigation. Most older methods judge this offline and ignore which way the robot is heading, even though heading changes the answer.
Their neural network reads fine shape data about the scene together with the robot's recent movement, producing smooth estimates of traversability that account for heading. Tests ran on simulated and real robots in varied settings, and the method beat older approaches on measured performance. Strikingly, a model trained only in a simulator predicted real-world traversability well with no real-world training data at all. The authors argue the approach improves path planning and scouting in hard terrain, and they released the code publicly.
Catalog record
| CREATOR | Fetullah Atas, Grzegorz Cielniak, Lars Grimstad |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-06904