Vision-Language Navigation with Coarse Maps and Semantic Landmarks

2026 - Present
Zhejiang University

This ongoing research explores how human-readable navigation cues can complement metric perception in environments where dense maps are unavailable or unnecessarily heavy.

• Studying how navigation signs, object landmarks, and scene-level semantic structures can serve as reliable anchors for language-guided localization and route reasoning.
• Exploring spatial-semantic representations that connect geometric perception, place recognition, and language-conditioned navigation decisions in map-light robotic systems.
• Extending previous research experience in LiDAR-based place recognition and loop closure detection toward more interactive, instruction-aware, and cognitively grounded autonomous navigation.