Ground Slow, Move Fast: A Dual-System Foundation Model for Generalizable Vision-and-Language Navigation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wei, Meng, Wan, Chenyang, Peng, Jiaqi, Yu, Xiqian, Yang, Yuqiang, Feng, Delin, Cai, Wenzhe, Zhu, Chenming, Wang, Tai, Pang, Jiangmiao, Liu, Xihui
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909950583767040
author Wei, Meng
Wan, Chenyang
Peng, Jiaqi
Yu, Xiqian
Yang, Yuqiang
Feng, Delin
Cai, Wenzhe
Zhu, Chenming
Wang, Tai
Pang, Jiangmiao
Liu, Xihui
author_facet Wei, Meng
Wan, Chenyang
Peng, Jiaqi
Yu, Xiqian
Yang, Yuqiang
Feng, Delin
Cai, Wenzhe
Zhu, Chenming
Wang, Tai
Pang, Jiangmiao
Liu, Xihui
contents While recent large vision-language models (VLMs) have improved generalization in vision-language navigation (VLN), existing methods typically rely on end-to-end pipelines that map vision-language inputs directly to short-horizon discrete actions. Such designs often produce fragmented motions, incur high latency, and struggle with real-world challenges like dynamic obstacle avoidance. We propose DualVLN, the first dual-system VLN foundation model that synergistically integrates high-level reasoning with low-level action execution. System 2, a VLM-based global planner, "grounds slowly" by predicting mid-term waypoint goals via image-grounded reasoning. System 1, a lightweight, multi-modal conditioning Diffusion Transformer policy, "moves fast" by leveraging both explicit pixel goals and latent features from System 2 to generate smooth and accurate trajectories. The dual-system design enables robust real-time control and adaptive local decision-making in complex, dynamic environments. By decoupling training, the VLM retains its generalization, while System 1 achieves interpretable and effective local navigation. DualVLN outperforms prior methods across all VLN benchmarks and real-world experiments demonstrate robust long-horizon planning and real-time adaptability in dynamic environments.
format Preprint
id arxiv_https___arxiv_org_abs_2512_08186
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Ground Slow, Move Fast: A Dual-System Foundation Model for Generalizable Vision-and-Language Navigation
Wei, Meng
Wan, Chenyang
Peng, Jiaqi
Yu, Xiqian
Yang, Yuqiang
Feng, Delin
Cai, Wenzhe
Zhu, Chenming
Wang, Tai
Pang, Jiangmiao
Liu, Xihui
Robotics
While recent large vision-language models (VLMs) have improved generalization in vision-language navigation (VLN), existing methods typically rely on end-to-end pipelines that map vision-language inputs directly to short-horizon discrete actions. Such designs often produce fragmented motions, incur high latency, and struggle with real-world challenges like dynamic obstacle avoidance. We propose DualVLN, the first dual-system VLN foundation model that synergistically integrates high-level reasoning with low-level action execution. System 2, a VLM-based global planner, "grounds slowly" by predicting mid-term waypoint goals via image-grounded reasoning. System 1, a lightweight, multi-modal conditioning Diffusion Transformer policy, "moves fast" by leveraging both explicit pixel goals and latent features from System 2 to generate smooth and accurate trajectories. The dual-system design enables robust real-time control and adaptive local decision-making in complex, dynamic environments. By decoupling training, the VLM retains its generalization, while System 1 achieves interpretable and effective local navigation. DualVLN outperforms prior methods across all VLN benchmarks and real-world experiments demonstrate robust long-horizon planning and real-time adaptability in dynamic environments.
title Ground Slow, Move Fast: A Dual-System Foundation Model for Generalizable Vision-and-Language Navigation
topic Robotics
url https://arxiv.org/abs/2512.08186