DA-Nav: Direction-Aware City-Scale Vision-Language Navigation

2026-07-13

AI Summary

This work addresses the reliance of city-scale outdoor navigation on high-definition maps or costly supervisory signals by proposing a novel direction-aware vision-language navigation paradigm that operates solely with commercial navigation instructions. The approach reformulates the task as a discrete spatial localization problem in the egocentric image plane, integrating vision-language spatial grounding, chain-of-thought reasoning, and a discrete grid-based action space to enable end-to-end closed-loop control for deviation assessment, action prediction, and trajectory recovery. The study introduces ReDA, the first dataset featuring directional instructions paired with recovery trajectories, and achieves a state-of-the-art 56.16% success rate in unseen urban CARLA environments. Notably, the method generalizes to real-world complex settings without fine-tuning, enabling stable kilometer-scale navigation for both quadrupedal and humanoid robots.

Abstract

City-scale outdoor navigation is currently hindered by the heavy reliance on dense maps or costly navigation supervision. In this work, we introduce a novel paradigm for leveraging directional instructions from commercial navigation tools (e.g., Google Maps). To bridge the gap between commercial instructions and executable navigation actions, while mitigating long-horizon error accumulation through robust trajectory recovery, we propose DA-Nav, a Direction-Aware vision-language Navigation framework that reformulates navigation as a discrete spatial grounding problem on the egocentric 2D image plane. To achieve trajectory recovery, DA-Nav employs a Chain-of-Thought (CoT) reasoning process encompassing deviation assessment, action prediction, and target grid selection. We further introduce ReDA, a dataset that provides direction-aware instructions and recovery trajectories to enhance spatial grounding and support CoT recovery reasoning. Extensive experiments in CARLA demonstrate that DA-Nav achieves a high success rate of 56.16% in unseen urban environments, outperforming existing State-of-The-Art (SoTA) methods while maintaining a substantially stronger recovery capability. Furthermore, without fine-tuning, DA-Nav seamlessly adapts to both quadruped and humanoid robots, enabling stable kilometer-scale closed-loop outdoor navigation in complex real world environments.

Problem

• city-scale navigation
• vision-language navigation
• directional instructions
• trajectory recovery
• spatial grounding

Innovation

• Direction-Aware Navigation
• Vision-Language Navigation
• Chain-of-Thought Reasoning
• Spatial Grounding
• Trajectory Recovery