VertiFormer : A Data-Efficient Multi-Task Transformer for Off-Road Robot Mobility
Mohammad Nazeri 1 , Anuj Pokhrel 1 , Alexandyr Card 1 , Aniket Datar 1 , Garrett Warnell 2 3 and Xuesu Xiao 1 · 2025
Read VertiFormer : A Data-Efficient Multi-Task Transformer for Off-Road Robot Mobility by Mohammad Nazeri 1 , Anuj Pokhrel 1 , Alexandyr Card 1 , Aniket Datar 1 , Garrett Warnell 2 3 and Xuesu Xiao 1 (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-2502-00543.txt.
Description
Mohammad Nazeri, Anuj Pokhrel, and their co-authors built VertiFormer, a 2025 transformer model that helps off road robots move over rough ground. Robots must grasp how wheels and terrain interact on steep, rugged land. Unlike text or image fields, real off road driving data are scarce and costly to gather with physical robots.
VertiFormer learned from only one hour of driving data. A new masked training scheme teaches it to predict the next pose, the next action, and the next patch of terrain at once. That single training supports many tasks at once, including forward and inverse models of how the robot moves. The design targets terrain that is not just rough but sharply vertical in places.
Catalog record
| CREATOR | Mohammad Nazeri 1 , Anuj Pokhrel 1 , Alexandyr Card 1 , Aniket Datar 1 , Garrett Warnell 2 3 and Xuesu Xiao 1 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-00543