Learning to Act Robustly with View-Invariant Latent Actions
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909982621958144 |
|---|---|
| author | Jeong, Youngjoon Chun, Junha Kim, Taesup |
| author_facet | Jeong, Youngjoon Chun, Junha Kim, Taesup |
| contents | Vision-based robotic policies often struggle with even minor viewpoint changes, underscoring the need for view-invariant visual representations. This challenge becomes more pronounced in real-world settings, where viewpoint variability is unavoidable and can significantly disrupt policy performance. Existing methods typically learn invariance from multi-view observations at the scene level, but such approaches rely on visual appearance and fail to incorporate the physical dynamics essential for robust generalization. We propose View-Invariant Latent Action (VILA), which models a latent action capturing transition patterns across trajectories to learn view-invariant representations grounded in physical dynamics. VILA aligns these latent actions across viewpoints using an action-guided objective based on ground-truth action sequences. Experiments in both simulation and the real world show that VILA-based policies generalize effectively to unseen viewpoints and transfer well to new tasks, establishing VILA as a strong pretraining framework that improves robustness and downstream learning performance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_02994 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Learning to Act Robustly with View-Invariant Latent Actions Jeong, Youngjoon Chun, Junha Kim, Taesup Robotics Artificial Intelligence Machine Learning Vision-based robotic policies often struggle with even minor viewpoint changes, underscoring the need for view-invariant visual representations. This challenge becomes more pronounced in real-world settings, where viewpoint variability is unavoidable and can significantly disrupt policy performance. Existing methods typically learn invariance from multi-view observations at the scene level, but such approaches rely on visual appearance and fail to incorporate the physical dynamics essential for robust generalization. We propose View-Invariant Latent Action (VILA), which models a latent action capturing transition patterns across trajectories to learn view-invariant representations grounded in physical dynamics. VILA aligns these latent actions across viewpoints using an action-guided objective based on ground-truth action sequences. Experiments in both simulation and the real world show that VILA-based policies generalize effectively to unseen viewpoints and transfer well to new tasks, establishing VILA as a strong pretraining framework that improves robustness and downstream learning performance. |
| title | Learning to Act Robustly with View-Invariant Latent Actions |
| topic | Robotics Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2601.02994 |