FloNa: Floor Plan Guided Embodied Visual Navigation

TL;DR

FloNa introduces a new embodied visual navigation task that leverages floor plans to improve navigation efficiency and accuracy, addressing challenges of spatial inconsistency and modality alignment with a diffusion policy framework.

Contribution

This work pioneers the integration of floor plan prior knowledge into embodied visual navigation and proposes FloDiff, a diffusion-based policy with localization for better scene alignment.

Findings

FloNa significantly improves navigation performance in unfamiliar scenes.

The framework effectively handles spatial and modality discrepancies.

Extensive experiments validate the approach's efficiency and robustness.

Abstract

Humans naturally rely on floor plans to navigate in unfamiliar environments, as they are readily available, reliable, and provide rich geometrical guidance. However, existing visual navigation settings overlook this valuable prior knowledge, leading to limited efficiency and accuracy. To eliminate this gap, we introduce a novel navigation task: Floor Plan Visual Navigation (FloNa), the first attempt to incorporate floor plan into embodied visual navigation. While the floor plan offers significant advantages, two key challenges emerge: (1) handling the spatial inconsistency between the floor plan and the actual scene layout for collision-free navigation, and (2) aligning observed images with the floor plan sketch despite their distinct modalities. To address these challenges, we propose FloDiff, a novel diffusion policy framework incorporating a localization module to facilitate alignment between the current observation and the floor plan. We further collect $20k$ navigation episodes across $117$ scenes in the iGibson simulator to support the training and evaluation. Extensive experiments demonstrate the effectiveness and efficiency of our framework in unfamiliar scenes using floor plan knowledge. Project website: https://gauleejx.github.io/flona/.