Can I read Scaling Cross-Embodied Learning: One Policy For Manipulation, Navigation, Locomotion and Aviation on EtoBox?

Scaling Cross-Embodied Learning: One Policy For Manipulation, Navigation, Locomotion and Aviation by xiangrongraav is a document available to read on EtoBox.

What is Scaling Cross-Embodied Learning: One Policy For Manipulation, Navigation, Locomotion and Aviation about?

The document introduces CrossFormer, a transformer-based policy designed for cross-embodied robot learning, capable of controlling various robotic platforms using a single policy trained on 900K trajectories from 20 different embodiments. This approach allows for better generalization and robustness by leveraging diverse datasets without the need for manual alignment of observation or action spaces. Extensive experiments demonstrate that CrossFormer matches or exceeds the performance of specialized policies

Author

xiangrongraav

Language

EN