HRO: Hierarchical Room-to-Object Framework for Zero-Shot Object Goal Navigation with Large Language Models

arXiv cs.AI · 18 hours ago

Researchers developed HRO, a framework that uses large language models to guide robots in navigating to unknown objects in unfamiliar environments by modeling hierarchical spatial reasoning from rooms to specific objects. The method was evaluated on Gibson and HM3D datasets and achieved higher success rates and generalization compared to existing LLM-based navigation approaches. The framework improves navigation accuracy by applying coarse-to-fine exploration rather than treating object localization as a flat reasoning task.