We propose a zero-shot object-goal navigation framework by constructing an online 3D scene graph to prompt LLMs. Our method can be directly applied to different kinds of scenes and categories without ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results