Trajectory-Diversity-Driven Robust Vision-and-Language Navigation
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911519872122880 |
|---|---|
| author | Li, Jiangyang Wan, Cong Dong, SongLin Ding, Chenhao Wang, Qiang Ma, Zhiheng Gong, Yihong |
| author_facet | Li, Jiangyang Wan, Cong Dong, SongLin Ding, Chenhao Wang, Qiang Ma, Zhiheng Gong, Yihong |
| contents | Vision-and-Language Navigation (VLN) requires agents to navigate photo-realistic environments following natural language instructions. Current methods predominantly rely on imitation learning, which suffers from limited generalization and poor robustness to execution perturbations. We present NavGRPO, a reinforcement learning framework that learns goal-directed navigation policies through Group Relative Policy Optimization. By exploring diverse trajectories and optimizing via within-group performance comparisons, our method enables agents to distinguish effective strategies beyond expert paths without requiring additional value networks. Built on ScaleVLN, NavGRPO achieves superior robustness on R2R and REVERIE benchmarks with +3.0% and +1.71% SPL improvements in unseen environments. Under extreme early-stage perturbations, we demonstrate +14.89% SPL gain over the baseline, confirming that goal-directed RL training builds substantially more robust navigation policies. Code and models will be released. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_15370 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Trajectory-Diversity-Driven Robust Vision-and-Language Navigation Li, Jiangyang Wan, Cong Dong, SongLin Ding, Chenhao Wang, Qiang Ma, Zhiheng Gong, Yihong Computer Vision and Pattern Recognition Vision-and-Language Navigation (VLN) requires agents to navigate photo-realistic environments following natural language instructions. Current methods predominantly rely on imitation learning, which suffers from limited generalization and poor robustness to execution perturbations. We present NavGRPO, a reinforcement learning framework that learns goal-directed navigation policies through Group Relative Policy Optimization. By exploring diverse trajectories and optimizing via within-group performance comparisons, our method enables agents to distinguish effective strategies beyond expert paths without requiring additional value networks. Built on ScaleVLN, NavGRPO achieves superior robustness on R2R and REVERIE benchmarks with +3.0% and +1.71% SPL improvements in unseen environments. Under extreme early-stage perturbations, we demonstrate +14.89% SPL gain over the baseline, confirming that goal-directed RL training builds substantially more robust navigation policies. Code and models will be released. |
| title | Trajectory-Diversity-Driven Robust Vision-and-Language Navigation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.15370 |