Navi-Agent navigates unknown spaces using visual places not coordinates
Navi-Agent: Unlocalized Monocular Navigation Agent
RoboticsComputer Vision and Pattern Recognition
Summary
Navigating unfamiliar places using only a camera is tricky because it's hard to know exactly where you are without maps or GPS. The authors created Navi-Agent, a robot that remembers places it has seen as pictures linked together by how it moved between them. This way, it can figure out its location approximately, check how far it has progressed, and backtrack if needed without needing exact coordinates. They showed Navi-Agent works well on tests and real robots, keeping track of places visually instead of using traditional geometry.
What this means in practice
- •For robotics engineers: Develop robots that navigate complex new environments using only camera images, without relying on detailed maps or GPS.$Commercial implications: Enables navigation products for robots in unknown indoor or outdoor settings where GPS or mapping is unavailable or costly.
- •For machine learning practitioners: Build visual navigation systems for autonomous agents that verify progress and recover from errors through visual scene recognition instead of coordinate tracking.
Authors
Wenyuan Xie, Mengyang Hong, Yongzhong Wang, Yanbiao Ji, Yijin Zhou, Shaokai Wu, Shalayiding Sirejiding, Huayi Zhou, Yi-Chao Chen, Ma Ling, Yue Ding, Hongtao Lu
Abstract
Vision-Language Navigation in Continuous Environments (VLN-CE) requires an embodied agent to execute long-horizon instructions in unknown environments. Existing zero-shot VLN-CE systems typically maintain spatial states through geometric localization or coordinate-based representations. Recent geometry-constrained navigation removes depth and globally consistent coordinates, but maintaining persistent spatial awareness for place confirmation, progress verification, and recovery remains challenging. We present Navi-Agent, a zero-shot VLN-CE agent that constructs a coordinate-free spatial state from visual observations and executed motion histories. Navi-Agent organizes this state as a navigation topology, where nodes represent visual places and edges represent motion transitions. This representation enables observation-based approximate self-localization, task progress verification, and visual revisitation-based recovery. Navi-Agent performs closed-loop navigation by decomposing instructions into sub-goals, executing local visual navigation, and verifying visited places through the constructed spatial state. Experiments on zero-shot VLN-CE benchmark and real-world robot platforms show that Navi-Agent achieves state-of-the-art performance among geometry-constrained methods while remaining competitive with approaches relying on geometric localization.