Learning Modular Policy for Multi-Floor Object Navigation:A Factorized Framework for Diagnostic Study

What happened

arXiv:2610.06958v1 Announce Type: new Abstract: Object-goal navigation (ObjectNav) in multi-floor scenarios presents a challenge due to sparse rewards caused by long-horizon decision-making. Experiment results indicate that perception performance and stair climbing stability are the primary bottlenecks in multi-floor navigation.

To providing an effective initialization for Reinforcement Learning (RL), the lightweight intra-floor policy is learned by distilling the exploration logic of Visual Language Models (VLMs). arXiv is committed to these values and only works with partners that adhere to them.